Deep Learning Engineer
For engineers who want to design, train, scale, and deploy deep neural networks. You will be able to build vision, sequence, generative, and LLM systems and run them efficiently in production.
{"nodes":[{"id":"title","type":"title","position":{"x":0,"y":0},"data":{"label":"Deep Learning Engineer"},"width":1688,"height":61,"style":{"width":1688,"height":61},"measured":{"width":1688,"height":61},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":0}},{"id":"summary","type":"paragraph","position":{"x":404,"y":77},"data":{"label":"For engineers who want to design, train, scale, and deploy deep neural networks. You will be able to build vision, sequence, generative, and LLM systems and run them efficiently in production."},"width":880,"height":56,"style":{"width":880,"height":56},"measured":{"width":880,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":404,"y":77}},{"id":"meta","type":"paragraph","position":{"x":524,"y":147},"data":{"label":"10 STAGES · 40 TOPICS · 168 SUBTOPICS","style":{"fontSize":13,"fontFamily":"jetbrains","fontWeight":500,"color":"var(--color-fg-subtle)"}},"width":640,"height":21,"style":{"width":640,"height":21},"measured":{"width":640,"height":21},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":524,"y":147}},{"id":"stage-1","type":"section","position":{"x":0,"y":224},"data":{"label":"Prerequisites & Fundamentals","description":"Get fluent in Python, core math, and classical ML before touching neural networks.","number":1},"width":824,"height":937,"style":{"width":824,"height":937},"measured":{"width":824,"height":937},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":224}},{"id":"topic-1-1","type":"topic","position":{"x":28,"y":345},"data":{"label":"Programming & Software Engineering","number":"1.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":345}},{"id":"sub-1-1-1","type":"subtopic","position":{"x":28,"y":413},"data":{"label":"Python Programming","description":"Data Structures · OOP · Typing · Async"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":413}},{"id":"sub-1-1-2","type":"subtopic","position":{"x":288,"y":413},"data":{"label":"Data Manipulation","description":"NumPy · Pandas · Polars"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":413}},{"id":"sub-1-1-3","type":"subtopic","position":{"x":548,"y":413},"data":{"label":"Linux & Command Line"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":413}},{"id":"sub-1-1-4","type":"subtopic","position":{"x":28,"y":510},"data":{"label":"Version Control","description":"Git · GitHub · GitLab"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":510}},{"id":"sub-1-1-5","type":"subtopic","position":{"x":288,"y":510},"data":{"label":"Clean Code & Design Patterns"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":510}},{"id":"topic-1-2","type":"topic","position":{"x":28,"y":605},"data":{"label":"Applied Mathematics","number":"1.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":605}},{"id":"sub-1-2-1","type":"subtopic","position":{"x":28,"y":673},"data":{"label":"Linear Algebra","description":"Vectors · Matrices · Eigenvalues · SVD"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":673}},{"id":"sub-1-2-2","type":"subtopic","position":{"x":288,"y":673},"data":{"label":"Calculus","description":"Derivatives · Partial Derivatives · Chain Rule"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":673}},{"id":"sub-1-2-3","type":"subtopic","position":{"x":548,"y":673},"data":{"label":"Probability & Statistics","description":"Distributions · Bayes' Theorem · MLE"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":673}},{"id":"sub-1-2-4","type":"subtopic","position":{"x":28,"y":770},"data":{"label":"Information Theory","description":"Entropy · Cross-Entropy · KL Divergence"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":770}},{"id":"topic-1-3","type":"topic","position":{"x":28,"y":883},"data":{"label":"Machine Learning Basics","number":"1.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":883}},{"id":"sub-1-3-1","type":"subtopic","position":{"x":28,"y":951},"data":{"label":"Supervised vs Unsupervised Learning"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":951}},{"id":"sub-1-3-2","type":"subtopic","position":{"x":288,"y":951},"data":{"label":"Linear & Logistic Regression"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":951}},{"id":"sub-1-3-3","type":"subtopic","position":{"x":548,"y":951},"data":{"label":"Classical Models","description":"SVM · Decision Trees · k-NN"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":951}},{"id":"sub-1-3-4","type":"subtopic","position":{"x":28,"y":1030},"data":{"label":"Ensemble Methods","description":"Random Forest · Gradient Boosting · XGBoost"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1030}},{"id":"sub-1-3-5","type":"subtopic","position":{"x":288,"y":1030},"data":{"label":"Bias-Variance & Overfitting"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1030}},{"id":"sub-1-3-6","type":"subtopic","position":{"x":548,"y":1030},"data":{"label":"Evaluation Metrics","description":"Accuracy · Precision · Recall · F1 · ROC-AUC"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":1030}},{"id":"stage-2","type":"section","position":{"x":864,"y":224},"data":{"label":"Neural Network Foundations","description":"Understand how neural networks learn, from forward passes to stable, well-tuned training.","number":2},"width":824,"height":937,"style":{"width":824,"height":937},"measured":{"width":824,"height":937},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":864,"y":224}},{"id":"topic-2-1","type":"topic","position":{"x":892,"y":345},"data":{"label":"Artificial Neural Networks","number":"2.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":345}},{"id":"sub-2-1-1","type":"subtopic","position":{"x":892,"y":413},"data":{"label":"Perceptrons & Multi-Layer Perceptrons (MLPs)"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":413}},{"id":"sub-2-1-2","type":"subtopic","position":{"x":1152,"y":413},"data":{"label":"Activation Functions","description":"Sigmoid · Tanh · ReLU · GELU · Swish"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":413}},{"id":"sub-2-1-3","type":"subtopic","position":{"x":1412,"y":413},"data":{"label":"Forward Propagation & Backpropagation"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":413}},{"id":"sub-2-1-4","type":"subtopic","position":{"x":892,"y":510},"data":{"label":"Computational Graphs & Autodiff"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":510}},{"id":"topic-2-2","type":"topic","position":{"x":892,"y":603},"data":{"label":"Training Neural Networks","number":"2.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":603}},{"id":"sub-2-2-1","type":"subtopic","position":{"x":892,"y":671},"data":{"label":"Loss Functions","description":"MSE · Cross-Entropy · Huber · Focal"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":671}},{"id":"sub-2-2-2","type":"subtopic","position":{"x":1152,"y":671},"data":{"label":"Weight Initialization","description":"Xavier · He · Orthogonal"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":671}},{"id":"sub-2-2-3","type":"subtopic","position":{"x":1412,"y":671},"data":{"label":"Normalization","description":"BatchNorm · LayerNorm · GroupNorm · RMSNorm"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":671}},{"id":"sub-2-2-4","type":"subtopic","position":{"x":892,"y":768},"data":{"label":"Regularization","description":"L1 · L2 · Dropout · Early Stopping · Weight Decay"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":768}},{"id":"sub-2-2-5","type":"subtopic","position":{"x":1152,"y":768},"data":{"label":"Debugging Training","description":"Vanishing Gradients · Gradient Clipping · Overfit One Batch"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":768}},{"id":"topic-2-3","type":"topic","position":{"x":892,"y":881},"data":{"label":"Optimizers & Schedules","number":"2.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":881}},{"id":"sub-2-3-1","type":"subtopic","position":{"x":892,"y":949},"data":{"label":"Gradient Descent Variants","description":"Batch · Mini-batch · SGD · Momentum · Nesterov"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":949}},{"id":"sub-2-3-2","type":"subtopic","position":{"x":1152,"y":949},"data":{"label":"Adaptive Optimizers","description":"AdaGrad · RMSprop · Adam · AdamW"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":949}},{"id":"sub-2-3-3","type":"subtopic","position":{"x":1412,"y":949},"data":{"label":"Large-Batch & Modern Optimizers","description":"LAMB · Lion · Muon"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":949}},{"id":"sub-2-3-4","type":"subtopic","position":{"x":892,"y":1048},"data":{"label":"Learning Rate Schedules","description":"Warmup · Cosine Annealing · One-Cycle"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1048}},{"id":"stage-3","type":"section","position":{"x":0,"y":1201},"data":{"label":"Deep Learning Frameworks","description":"Implement, track, and tune models in PyTorch, with working knowledge of Keras and JAX.","number":3},"width":824,"height":962,"style":{"width":824,"height":962},"measured":{"width":824,"height":962},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":1201}},{"id":"topic-3-1","type":"topic","position":{"x":28,"y":1322},"data":{"label":"PyTorch","number":"3.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1322}},{"id":"sub-3-1-1","type":"subtopic","position":{"x":28,"y":1390},"data":{"label":"Tensors & Autograd"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1390}},{"id":"sub-3-1-2","type":"subtopic","position":{"x":288,"y":1390},"data":{"label":"Building nn.Module Classes"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1390}},{"id":"sub-3-1-3","type":"subtopic","position":{"x":548,"y":1390},"data":{"label":"Datasets & DataLoaders"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":1390}},{"id":"sub-3-1-4","type":"subtopic","position":{"x":28,"y":1448},"data":{"label":"Training Loops & Checkpointing"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1448}},{"id":"sub-3-1-5","type":"subtopic","position":{"x":288,"y":1448},"data":{"label":"Higher-Level Wrappers","description":"PyTorch Lightning · Hugging Face Accelerate"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1448}},{"id":"topic-3-2","type":"topic","position":{"x":28,"y":1561},"data":{"label":"TensorFlow & Keras","number":"3.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1561}},{"id":"sub-3-2-1","type":"subtopic","position":{"x":28,"y":1629},"data":{"label":"Eager Execution vs Graph Mode (tf.function)"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1629}},{"id":"sub-3-2-2","type":"subtopic","position":{"x":288,"y":1629},"data":{"label":"Keras 3 APIs","description":"Sequential · Functional · Subclassing"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1629}},{"id":"sub-3-2-3","type":"subtopic","position":{"x":548,"y":1629},"data":{"label":"Input Pipelines (tf.data)"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":1629}},{"id":"topic-3-3","type":"topic","position":{"x":28,"y":1742},"data":{"label":"JAX Ecosystem","number":"3.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1742}},{"id":"sub-3-3-1","type":"subtopic","position":{"x":28,"y":1810},"data":{"label":"JAX Transformations","description":"jit · grad · vmap · shard_map"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1810}},{"id":"sub-3-3-2","type":"subtopic","position":{"x":288,"y":1810},"data":{"label":"Neural Network Libraries","description":"Flax NNX · Equinox"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1810}},{"id":"sub-3-3-3","type":"subtopic","position":{"x":548,"y":1810},"data":{"label":"Optimization with Optax"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":1810}},{"id":"topic-3-4","type":"topic","position":{"x":28,"y":1905},"data":{"label":"Experiment Tracking & Tuning","number":"3.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1905}},{"id":"sub-3-4-1","type":"subtopic","position":{"x":28,"y":1973},"data":{"label":"Experiment Tracking","description":"TensorBoard · Weights & Biases · MLflow"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":1973}},{"id":"sub-3-4-2","type":"subtopic","position":{"x":288,"y":1973},"data":{"label":"Grid & Random Search"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":1973}},{"id":"sub-3-4-3","type":"subtopic","position":{"x":548,"y":1973},"data":{"label":"Bayesian Optimization","description":"Optuna · Ray Tune"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":1973}},{"id":"sub-3-4-4","type":"subtopic","position":{"x":28,"y":2070},"data":{"label":"Population Based Training (PBT)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2070}},{"id":"stage-4","type":"section","position":{"x":864,"y":1201},"data":{"label":"CNNs & Computer Vision","description":"Build convolutional networks for classification, detection, and segmentation.","number":4},"width":824,"height":962,"style":{"width":824,"height":962},"measured":{"width":824,"height":962},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":864,"y":1201}},{"id":"topic-4-1","type":"topic","position":{"x":892,"y":1322},"data":{"label":"Convolution Fundamentals","number":"4.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1322}},{"id":"sub-4-1-1","type":"subtopic","position":{"x":892,"y":1390},"data":{"label":"Convolutions, Padding & Strides"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1390}},{"id":"sub-4-1-2","type":"subtopic","position":{"x":1152,"y":1390},"data":{"label":"Pooling & Receptive Fields"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":1390}},{"id":"sub-4-1-3","type":"subtopic","position":{"x":1412,"y":1390},"data":{"label":"Data Augmentation","description":"Flips · Crops · Mixup · CutMix"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":1390}},{"id":"topic-4-2","type":"topic","position":{"x":892,"y":1485},"data":{"label":"CNN Architectures","number":"4.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1485}},{"id":"sub-4-2-1","type":"subtopic","position":{"x":892,"y":1553},"data":{"label":"Classic Architectures","description":"LeNet · AlexNet · VGG"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1553}},{"id":"sub-4-2-2","type":"subtopic","position":{"x":1152,"y":1553},"data":{"label":"Modern Architectures","description":"ResNet · Inception · EfficientNet · ConvNeXt"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":1553}},{"id":"sub-4-2-3","type":"subtopic","position":{"x":1412,"y":1553},"data":{"label":"Transfer Learning & Fine-Tuning","description":"timm · torchvision"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":1553}},{"id":"topic-4-3","type":"topic","position":{"x":892,"y":1668},"data":{"label":"Object Detection","number":"4.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1668}},{"id":"sub-4-3-1","type":"subtopic","position":{"x":892,"y":1736},"data":{"label":"Two-Stage Detectors","description":"R-CNN · Faster R-CNN"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1736}},{"id":"sub-4-3-2","type":"subtopic","position":{"x":1152,"y":1736},"data":{"label":"One-Stage Detectors","description":"YOLO Family · SSD · RetinaNet"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":1736}},{"id":"sub-4-3-3","type":"subtopic","position":{"x":1412,"y":1736},"data":{"label":"Detection Metrics","description":"IoU · NMS · mAP"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":1736}},{"id":"topic-4-4","type":"topic","position":{"x":892,"y":1831},"data":{"label":"Image Segmentation","number":"4.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1831}},{"id":"sub-4-4-1","type":"subtopic","position":{"x":892,"y":1899},"data":{"label":"Semantic Segmentation","description":"FCN · U-Net · DeepLab"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":1899}},{"id":"sub-4-4-2","type":"subtopic","position":{"x":1152,"y":1899},"data":{"label":"Instance & Panoptic Segmentation","description":"Mask R-CNN · Mask2Former"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":1899}},{"id":"stage-5","type":"section","position":{"x":0,"y":2202},"data":{"label":"Sequence Models & Transformers","description":"Model sequential data with RNNs and master the Transformer that powers modern AI.","number":5},"width":824,"height":1541,"style":{"width":824,"height":1541},"measured":{"width":824,"height":1541},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":2202}},{"id":"topic-5-1","type":"topic","position":{"x":28,"y":2323},"data":{"label":"Recurrent Networks","number":"5.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2323}},{"id":"sub-5-1-1","type":"subtopic","position":{"x":28,"y":2391},"data":{"label":"Recurrent Neural Networks (RNNs)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2391}},{"id":"sub-5-1-2","type":"subtopic","position":{"x":288,"y":2391},"data":{"label":"LSTMs & GRUs"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":2391}},{"id":"sub-5-1-3","type":"subtopic","position":{"x":548,"y":2391},"data":{"label":"Bidirectional & Stacked RNNs"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":2391}},{"id":"sub-5-1-4","type":"subtopic","position":{"x":28,"y":2468},"data":{"label":"Seq2Seq with Attention","description":"Additive · Multiplicative"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2468}},{"id":"topic-5-2","type":"topic","position":{"x":28,"y":2563},"data":{"label":"Embeddings & Tokenization","number":"5.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2563}},{"id":"sub-5-2-1","type":"subtopic","position":{"x":28,"y":2631},"data":{"label":"Static Embeddings","description":"Word2Vec · GloVe · FastText"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2631}},{"id":"sub-5-2-2","type":"subtopic","position":{"x":288,"y":2631},"data":{"label":"Character-Level Embeddings"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":2631}},{"id":"sub-5-2-3","type":"subtopic","position":{"x":548,"y":2631},"data":{"label":"Contextual Embeddings (ELMo)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":2631}},{"id":"sub-5-2-4","type":"subtopic","position":{"x":28,"y":2710},"data":{"label":"Tokenization Algorithms","description":"BPE · WordPiece · SentencePiece · tiktoken"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2710}},{"id":"topic-5-3","type":"topic","position":{"x":28,"y":2823},"data":{"label":"Transformer Architecture","number":"5.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2823}},{"id":"sub-5-3-1","type":"subtopic","position":{"x":28,"y":2891},"data":{"label":"Self-Attention & Multi-Head Attention"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2891}},{"id":"sub-5-3-2","type":"subtopic","position":{"x":288,"y":2891},"data":{"label":"Positional Encodings","description":"Sinusoidal · Learned · RoPE · ALiBi"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":2891}},{"id":"sub-5-3-3","type":"subtopic","position":{"x":548,"y":2891},"data":{"label":"Residuals & Feed-Forward Layers"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":2891}},{"id":"sub-5-3-4","type":"subtopic","position":{"x":28,"y":2988},"data":{"label":"Encoder-Only Models","description":"BERT · RoBERTa · DeBERTa · ModernBERT"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":2988}},{"id":"sub-5-3-5","type":"subtopic","position":{"x":288,"y":2988},"data":{"label":"Decoder-Only Models","description":"GPT · Llama · Mistral · Qwen"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":2988}},{"id":"sub-5-3-6","type":"subtopic","position":{"x":548,"y":2988},"data":{"label":"Encoder-Decoder Models","description":"T5 · BART"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":2988}},{"id":"topic-5-4","type":"topic","position":{"x":28,"y":3101},"data":{"label":"State Space Models & Alternatives","number":"5.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":3101}},{"id":"sub-5-4-1","type":"subtopic","position":{"x":28,"y":3169},"data":{"label":"State Space Models","description":"S4 · Mamba · Mamba-2"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":3169}},{"id":"sub-5-4-2","type":"subtopic","position":{"x":288,"y":3169},"data":{"label":"Hybrid Architectures (Jamba)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":3169}},{"id":"sub-5-4-3","type":"subtopic","position":{"x":548,"y":3169},"data":{"label":"RWKV & Linear Attention"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":3169}},{"id":"stage-6","type":"section","position":{"x":864,"y":2202},"data":{"label":"Advanced Architectures & Generative Models","description":"Explore modern vision, self-supervised, generative, graph, and reinforcement learning models.","number":6},"width":824,"height":1541,"style":{"width":824,"height":1541},"measured":{"width":824,"height":1541},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":864,"y":2202}},{"id":"topic-6-1","type":"topic","position":{"x":892,"y":2323},"data":{"label":"Advanced Computer Vision","number":"6.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2323}},{"id":"sub-6-1-1","type":"subtopic","position":{"x":892,"y":2391},"data":{"label":"Vision Transformers","description":"ViT · Swin · DeiT"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2391}},{"id":"sub-6-1-2","type":"subtopic","position":{"x":1152,"y":2391},"data":{"label":"Transformer Detectors","description":"DETR · RT-DETR"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":2391}},{"id":"sub-6-1-3","type":"subtopic","position":{"x":1412,"y":2391},"data":{"label":"Promptable Segmentation","description":"SAM · SAM 2"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":2391}},{"id":"sub-6-1-4","type":"subtopic","position":{"x":892,"y":2470},"data":{"label":"3D Vision","description":"NeRF · Gaussian Splatting · Point Clouds"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2470}},{"id":"sub-6-1-5","type":"subtopic","position":{"x":1152,"y":2470},"data":{"label":"Video Understanding","description":"Action Recognition · VideoMAE"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":2470}},{"id":"topic-6-2","type":"topic","position":{"x":892,"y":2583},"data":{"label":"Self-Supervised Learning","number":"6.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2583}},{"id":"sub-6-2-1","type":"subtopic","position":{"x":892,"y":2651},"data":{"label":"Contrastive Learning","description":"SimCLR · MoCo · BYOL"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2651}},{"id":"sub-6-2-2","type":"subtopic","position":{"x":1152,"y":2651},"data":{"label":"Masked Image Modeling","description":"MAE · BEiT"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":2651}},{"id":"sub-6-2-3","type":"subtopic","position":{"x":1412,"y":2651},"data":{"label":"Self-Distillation","description":"DINO · DINOv2"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":2651}},{"id":"topic-6-3","type":"topic","position":{"x":892,"y":2746},"data":{"label":"Autoencoders & GANs","number":"6.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2746}},{"id":"sub-6-3-1","type":"subtopic","position":{"x":892,"y":2814},"data":{"label":"Autoencoders","description":"Vanilla · Sparse · Denoising"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2814}},{"id":"sub-6-3-2","type":"subtopic","position":{"x":1152,"y":2814},"data":{"label":"Variational Autoencoders (VAEs)"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":2814}},{"id":"sub-6-3-3","type":"subtopic","position":{"x":1412,"y":2814},"data":{"label":"GAN Fundamentals","description":"Generator · Discriminator · Adversarial Loss"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":2814}},{"id":"sub-6-3-4","type":"subtopic","position":{"x":892,"y":2911},"data":{"label":"GAN Variants","description":"DCGAN · WGAN · StyleGAN"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":2911}},{"id":"sub-6-3-5","type":"subtopic","position":{"x":1152,"y":2911},"data":{"label":"Image-to-Image Translation","description":"Pix2Pix · CycleGAN · Style Transfer"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":2911}},{"id":"topic-6-4","type":"topic","position":{"x":892,"y":3024},"data":{"label":"Diffusion & Flow Models","number":"6.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3024}},{"id":"sub-6-4-1","type":"subtopic","position":{"x":892,"y":3092},"data":{"label":"Diffusion Fundamentals","description":"DDPM · DDIM · Noise Schedules"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3092}},{"id":"sub-6-4-2","type":"subtopic","position":{"x":1152,"y":3092},"data":{"label":"Classifier-Free Guidance"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":3092}},{"id":"sub-6-4-3","type":"subtopic","position":{"x":1412,"y":3092},"data":{"label":"Latent Diffusion","description":"Stable Diffusion · Flux"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":3092}},{"id":"sub-6-4-4","type":"subtopic","position":{"x":892,"y":3171},"data":{"label":"Diffusion Transformers (DiT)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3171}},{"id":"sub-6-4-5","type":"subtopic","position":{"x":1152,"y":3171},"data":{"label":"Flow Matching & Rectified Flow"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":3171}},{"id":"topic-6-5","type":"topic","position":{"x":892,"y":3264},"data":{"label":"Graph Neural Networks","number":"6.5"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3264}},{"id":"sub-6-5-1","type":"subtopic","position":{"x":892,"y":3332},"data":{"label":"Graph Convolutional Networks (GCNs)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3332}},{"id":"sub-6-5-2","type":"subtopic","position":{"x":1152,"y":3332},"data":{"label":"Message Passing Neural Networks (MPNNs)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":3332}},{"id":"sub-6-5-3","type":"subtopic","position":{"x":1412,"y":3332},"data":{"label":"Graph Attention Networks (GATs)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":3332}},{"id":"sub-6-5-4","type":"subtopic","position":{"x":892,"y":3409},"data":{"label":"GNN Libraries (PyTorch Geometric)"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3409}},{"id":"topic-6-6","type":"topic","position":{"x":892,"y":3502},"data":{"label":"Deep Reinforcement Learning","number":"6.6"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3502}},{"id":"sub-6-6-1","type":"subtopic","position":{"x":892,"y":3570},"data":{"label":"MDPs & Q-Learning"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3570}},{"id":"sub-6-6-2","type":"subtopic","position":{"x":1152,"y":3570},"data":{"label":"Deep Q-Networks (DQN)"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":3570}},{"id":"sub-6-6-3","type":"subtopic","position":{"x":1412,"y":3570},"data":{"label":"Policy Gradients & Actor-Critic","description":"PPO · SAC"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":3570}},{"id":"sub-6-6-4","type":"subtopic","position":{"x":892,"y":3669},"data":{"label":"Offline & Inverse RL"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3669}},{"id":"stage-7","type":"section","position":{"x":0,"y":3783},"data":{"label":"Hardware & Distributed Training","description":"Understand accelerators and train large models efficiently across many GPUs.","number":7},"width":824,"height":1186,"style":{"width":824,"height":1186},"measured":{"width":824,"height":1186},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":3783}},{"id":"topic-7-1","type":"topic","position":{"x":28,"y":3904},"data":{"label":"Hardware Accelerators","number":"7.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":3904}},{"id":"sub-7-1-1","type":"subtopic","position":{"x":28,"y":3972},"data":{"label":"GPU Fundamentals","description":"CUDA · cuDNN · ROCm"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":3972}},{"id":"sub-7-1-2","type":"subtopic","position":{"x":288,"y":3972},"data":{"label":"GPU Architecture","description":"Tensor Cores · HBM · NVLink"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":3972}},{"id":"sub-7-1-3","type":"subtopic","position":{"x":548,"y":3972},"data":{"label":"TPUs & Custom ASICs","description":"Google TPU · AWS Trainium · AWS Inferentia"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":3972}},{"id":"sub-7-1-4","type":"subtopic","position":{"x":28,"y":4069},"data":{"label":"Inference Chips","description":"NPUs · Groq LPU"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4069}},{"id":"topic-7-2","type":"topic","position":{"x":28,"y":4164},"data":{"label":"GPU Programming & Profiling","number":"7.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4164}},{"id":"sub-7-2-1","type":"subtopic","position":{"x":28,"y":4232},"data":{"label":"CUDA Kernel Basics"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4232}},{"id":"sub-7-2-2","type":"subtopic","position":{"x":288,"y":4232},"data":{"label":"Custom Kernels with Triton"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":4232}},{"id":"sub-7-2-3","type":"subtopic","position":{"x":548,"y":4232},"data":{"label":"Profiling","description":"PyTorch Profiler · Nsight Systems"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":4232}},{"id":"topic-7-3","type":"topic","position":{"x":28,"y":4345},"data":{"label":"Efficient Training","number":"7.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4345}},{"id":"sub-7-3-1","type":"subtopic","position":{"x":28,"y":4413},"data":{"label":"Mixed Precision","description":"FP16 · BF16 · FP8"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4413}},{"id":"sub-7-3-2","type":"subtopic","position":{"x":288,"y":4413},"data":{"label":"Gradient Checkpointing & Accumulation"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":4413}},{"id":"sub-7-3-3","type":"subtopic","position":{"x":548,"y":4413},"data":{"label":"Efficient Attention (FlashAttention)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":4413}},{"id":"sub-7-3-4","type":"subtopic","position":{"x":28,"y":4492},"data":{"label":"Graph Compilation","description":"torch.compile · XLA"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4492}},{"id":"topic-7-4","type":"topic","position":{"x":28,"y":4587},"data":{"label":"Distributed Training","number":"7.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4587}},{"id":"sub-7-4-1","type":"subtopic","position":{"x":28,"y":4655},"data":{"label":"Data Parallelism (DDP)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4655}},{"id":"sub-7-4-2","type":"subtopic","position":{"x":288,"y":4655},"data":{"label":"Model Parallelism","description":"Tensor · Pipeline"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":4655}},{"id":"sub-7-4-3","type":"subtopic","position":{"x":548,"y":4655},"data":{"label":"Sharded Training","description":"FSDP · DeepSpeed ZeRO"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":4655}},{"id":"sub-7-4-4","type":"subtopic","position":{"x":28,"y":4734},"data":{"label":"Context Parallelism (Ring Attention)"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":4734}},{"id":"sub-7-4-5","type":"subtopic","position":{"x":288,"y":4734},"data":{"label":"Distributed Tooling","description":"torchrun · Accelerate · Megatron-LM"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":4734}},{"id":"stage-8","type":"section","position":{"x":864,"y":3783},"data":{"label":"Large Language Models & Multimodal AI","description":"Pretrain, fine-tune, align, and apply LLMs and multimodal foundation models.","number":8},"width":824,"height":1186,"style":{"width":824,"height":1186},"measured":{"width":824,"height":1186},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":864,"y":3783}},{"id":"topic-8-1","type":"topic","position":{"x":892,"y":3904},"data":{"label":"LLM Foundations","number":"8.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3904}},{"id":"sub-8-1-1","type":"subtopic","position":{"x":892,"y":3972},"data":{"label":"Pretraining Objectives","description":"Causal LM · Masked LM"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":3972}},{"id":"sub-8-1-2","type":"subtopic","position":{"x":1152,"y":3972},"data":{"label":"Scaling Laws & Emergent Abilities"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":3972}},{"id":"sub-8-1-3","type":"subtopic","position":{"x":1412,"y":3972},"data":{"label":"Mixture of Experts (MoE)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":3972}},{"id":"sub-8-1-4","type":"subtopic","position":{"x":892,"y":4051},"data":{"label":"Long Context","description":"RoPE Scaling · YaRN"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4051}},{"id":"sub-8-1-5","type":"subtopic","position":{"x":1152,"y":4051},"data":{"label":"Reasoning Models","description":"Test-Time Compute · Verifiable Rewards"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4051}},{"id":"topic-8-2","type":"topic","position":{"x":892,"y":4164},"data":{"label":"Fine-Tuning & Alignment","number":"8.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4164}},{"id":"sub-8-2-1","type":"subtopic","position":{"x":892,"y":4232},"data":{"label":"Supervised Fine-Tuning (SFT)"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4232}},{"id":"sub-8-2-2","type":"subtopic","position":{"x":1152,"y":4232},"data":{"label":"Parameter-Efficient Fine-Tuning","description":"LoRA · QLoRA · DoRA"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4232}},{"id":"sub-8-2-3","type":"subtopic","position":{"x":1412,"y":4232},"data":{"label":"RLHF","description":"Reward Models · PPO · GRPO"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":4232}},{"id":"sub-8-2-4","type":"subtopic","position":{"x":892,"y":4331},"data":{"label":"Preference Optimization","description":"DPO · ORPO · KTO"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4331}},{"id":"sub-8-2-5","type":"subtopic","position":{"x":1152,"y":4331},"data":{"label":"Fine-Tuning Tools","description":"TRL · PEFT · Unsloth · Axolotl"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4331}},{"id":"topic-8-3","type":"topic","position":{"x":892,"y":4426},"data":{"label":"Building with LLMs","number":"8.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4426}},{"id":"sub-8-3-1","type":"subtopic","position":{"x":892,"y":4494},"data":{"label":"Prompting & In-Context Learning"},"width":248,"height":106,"style":{"width":248,"height":106},"measured":{"width":248,"height":106},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4494}},{"id":"sub-8-3-2","type":"subtopic","position":{"x":1152,"y":4494},"data":{"label":"Retrieval-Augmented Generation","description":"Embeddings · Vector Databases · Reranking"},"width":248,"height":106,"style":{"width":248,"height":106},"measured":{"width":248,"height":106},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4494}},{"id":"sub-8-3-3","type":"subtopic","position":{"x":1412,"y":4494},"data":{"label":"Agentic Workflows","description":"ReAct · LangGraph · LlamaIndex · MCP"},"width":248,"height":106,"style":{"width":248,"height":106},"measured":{"width":248,"height":106},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":4494}},{"id":"sub-8-3-4","type":"subtopic","position":{"x":892,"y":4612},"data":{"label":"LLM Evaluation","description":"Benchmarks · LLM-as-a-Judge"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4612}},{"id":"topic-8-4","type":"topic","position":{"x":892,"y":4707},"data":{"label":"Multimodal Models","number":"8.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4707}},{"id":"sub-8-4-1","type":"subtopic","position":{"x":892,"y":4775},"data":{"label":"Contrastive Image-Text Models","description":"CLIP · SigLIP"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4775}},{"id":"sub-8-4-2","type":"subtopic","position":{"x":1152,"y":4775},"data":{"label":"Vision-Language Models","description":"LLaVA · Qwen-VL · InternVL"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4775}},{"id":"sub-8-4-3","type":"subtopic","position":{"x":1412,"y":4775},"data":{"label":"Visual Question Answering (VQA)"},"width":248,"height":87,"style":{"width":248,"height":87},"measured":{"width":248,"height":87},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":4775}},{"id":"sub-8-4-4","type":"subtopic","position":{"x":892,"y":4874},"data":{"label":"Speech & Audio Models","description":"Whisper · WaveNet · VITS"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":4874}},{"id":"sub-8-4-5","type":"subtopic","position":{"x":1152,"y":4874},"data":{"label":"Video Generation","description":"Sora · Veo"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":4874}},{"id":"stage-9","type":"section","position":{"x":0,"y":5008},"data":{"label":"Inference Optimization & Production","description":"Compress, serve, and operate deep learning models reliably at scale and on the edge.","number":9},"width":824,"height":1283,"style":{"width":824,"height":1283},"measured":{"width":824,"height":1283},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":0,"y":5008}},{"id":"topic-9-1","type":"topic","position":{"x":28,"y":5129},"data":{"label":"Model Optimization","number":"9.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5129}},{"id":"sub-9-1-1","type":"subtopic","position":{"x":28,"y":5197},"data":{"label":"Quantization","description":"PTQ · QAT · GPTQ · AWQ · GGUF"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5197}},{"id":"sub-9-1-2","type":"subtopic","position":{"x":288,"y":5197},"data":{"label":"Pruning","description":"Magnitude · Structured · Unstructured"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5197}},{"id":"sub-9-1-3","type":"subtopic","position":{"x":548,"y":5197},"data":{"label":"Knowledge Distillation"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":5197}},{"id":"sub-9-1-4","type":"subtopic","position":{"x":28,"y":5294},"data":{"label":"KV Cache & PagedAttention"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5294}},{"id":"sub-9-1-5","type":"subtopic","position":{"x":288,"y":5294},"data":{"label":"Speculative Decoding"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5294}},{"id":"topic-9-2","type":"topic","position":{"x":28,"y":5368},"data":{"label":"Inference Runtimes & Edge","number":"9.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5368}},{"id":"sub-9-2-1","type":"subtopic","position":{"x":28,"y":5436},"data":{"label":"ONNX & ONNX Runtime"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5436}},{"id":"sub-9-2-2","type":"subtopic","position":{"x":288,"y":5436},"data":{"label":"NVIDIA Stack","description":"TensorRT · TensorRT-LLM"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5436}},{"id":"sub-9-2-3","type":"subtopic","position":{"x":548,"y":5436},"data":{"label":"Intel Hardware (OpenVINO)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":5436}},{"id":"sub-9-2-4","type":"subtopic","position":{"x":28,"y":5515},"data":{"label":"Mobile Frameworks","description":"LiteRT · Core ML · ExecuTorch"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5515}},{"id":"sub-9-2-5","type":"subtopic","position":{"x":288,"y":5515},"data":{"label":"Edge Devices","description":"Jetson Orin · Raspberry Pi"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5515}},{"id":"topic-9-3","type":"topic","position":{"x":28,"y":5610},"data":{"label":"Model Serving","number":"9.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5610}},{"id":"sub-9-3-1","type":"subtopic","position":{"x":28,"y":5678},"data":{"label":"REST APIs","description":"FastAPI · Flask"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5678}},{"id":"sub-9-3-2","type":"subtopic","position":{"x":288,"y":5678},"data":{"label":"gRPC & Protocol Buffers"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5678}},{"id":"sub-9-3-3","type":"subtopic","position":{"x":548,"y":5678},"data":{"label":"Inference Servers","description":"Triton · vLLM · SGLang · Ray Serve · BentoML"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":5678}},{"id":"sub-9-3-4","type":"subtopic","position":{"x":28,"y":5775},"data":{"label":"Dynamic Batching & Autoscaling"},"width":248,"height":65,"style":{"width":248,"height":65},"measured":{"width":248,"height":65},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5775}},{"id":"topic-9-4","type":"topic","position":{"x":28,"y":5868},"data":{"label":"MLOps Infrastructure","number":"9.4"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5868}},{"id":"sub-9-4-1","type":"subtopic","position":{"x":28,"y":5936},"data":{"label":"Docker & Container Best Practices"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":5936}},{"id":"sub-9-4-2","type":"subtopic","position":{"x":288,"y":5936},"data":{"label":"Kubernetes for ML","description":"GPU Scheduling · KServe"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":5936}},{"id":"sub-9-4-3","type":"subtopic","position":{"x":548,"y":5936},"data":{"label":"ML Pipelines & Lifecycle","description":"Kubeflow · MLflow"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":5936}},{"id":"sub-9-4-4","type":"subtopic","position":{"x":28,"y":6015},"data":{"label":"CI/CD for Machine Learning"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":6015}},{"id":"sub-9-4-5","type":"subtopic","position":{"x":288,"y":6015},"data":{"label":"Data Version Control (DVC)"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":6015}},{"id":"sub-9-4-6","type":"subtopic","position":{"x":548,"y":6015},"data":{"label":"Feature Stores","description":"Feast · Hopsworks"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":6015}},{"id":"topic-9-5","type":"topic","position":{"x":28,"y":6110},"data":{"label":"Monitoring & Maintenance","number":"9.5"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":6110}},{"id":"sub-9-5-1","type":"subtopic","position":{"x":28,"y":6178},"data":{"label":"Data Drift & Model Drift Detection"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":28,"y":6178}},{"id":"sub-9-5-2","type":"subtopic","position":{"x":288,"y":6178},"data":{"label":"Logging & Telemetry","description":"Prometheus · Grafana · OpenTelemetry"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":288,"y":6178}},{"id":"sub-9-5-3","type":"subtopic","position":{"x":548,"y":6178},"data":{"label":"Safe Rollouts","description":"A/B Tests · Shadow Deployments · Canary Releases"},"width":248,"height":85,"style":{"width":248,"height":85},"measured":{"width":248,"height":85},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":548,"y":6178}},{"id":"stage-10","type":"section","position":{"x":864,"y":5008},"data":{"label":"Capstone & Career","description":"Demonstrate end-to-end deep learning skills and prepare for engineering interviews.","number":10},"width":824,"height":1283,"style":{"width":824,"height":1283},"measured":{"width":824,"height":1283},"zIndex":-999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":864,"y":5008}},{"id":"topic-10-1","type":"topic","position":{"x":892,"y":5129},"data":{"label":"Portfolio Projects","number":"10.1"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5129}},{"id":"sub-10-1-1","type":"subtopic","position":{"x":892,"y":5197},"data":{"label":"Architectures from Scratch","description":"MLP · CNN · Transformer"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5197}},{"id":"sub-10-1-2","type":"subtopic","position":{"x":1152,"y":5197},"data":{"label":"End-to-End Vision System"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":5197}},{"id":"sub-10-1-3","type":"subtopic","position":{"x":1412,"y":5197},"data":{"label":"Fine-Tune & Deploy an LLM"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":5197}},{"id":"sub-10-1-4","type":"subtopic","position":{"x":892,"y":5276},"data":{"label":"Reproduce a Research Paper"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5276}},{"id":"sub-10-1-5","type":"subtopic","position":{"x":1152,"y":5276},"data":{"label":"Open Source Contributions","description":"PyTorch · Hugging Face"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":5276}},{"id":"topic-10-2","type":"topic","position":{"x":892,"y":5371},"data":{"label":"Research Skills","number":"10.2"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5371}},{"id":"sub-10-2-1","type":"subtopic","position":{"x":892,"y":5439},"data":{"label":"Reading Papers","description":"arXiv · Hugging Face Papers"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5439}},{"id":"sub-10-2-2","type":"subtopic","position":{"x":1152,"y":5439},"data":{"label":"Experiment Design & Ablations"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":5439}},{"id":"sub-10-2-3","type":"subtopic","position":{"x":1412,"y":5439},"data":{"label":"Technical Writing & Blogging"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":5439}},{"id":"topic-10-3","type":"topic","position":{"x":892,"y":5534},"data":{"label":"Job Preparation","number":"10.3"},"width":768,"height":56,"style":{"width":768,"height":56},"measured":{"width":768,"height":56},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5534}},{"id":"sub-10-3-1","type":"subtopic","position":{"x":892,"y":5602},"data":{"label":"Resume & Portfolio"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5602}},{"id":"sub-10-3-2","type":"subtopic","position":{"x":1152,"y":5602},"data":{"label":"Deep Learning Theory Interviews"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":5602}},{"id":"sub-10-3-3","type":"subtopic","position":{"x":1412,"y":5602},"data":{"label":"Coding Interviews","description":"Python · PyTorch · DSA"},"width":248,"height":67,"style":{"width":248,"height":67},"measured":{"width":248,"height":67},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1412,"y":5602}},{"id":"sub-10-3-4","type":"subtopic","position":{"x":892,"y":5681},"data":{"label":"ML System Design"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":892,"y":5681}},{"id":"sub-10-3-5","type":"subtopic","position":{"x":1152,"y":5681},"data":{"label":"Take-Home Assignments"},"width":248,"height":46,"style":{"width":248,"height":46},"measured":{"width":248,"height":46},"zIndex":999,"selected":false,"selectable":true,"focusable":true,"dragging":false,"resizing":false,"positionAbsolute":{"x":1152,"y":5681}}],"edges":[]}
01 Prerequisites & Fundamentals Get fluent in Python, core math, and classical ML before touching neural networks.
02 Neural Network Foundations Understand how neural networks learn, from forward passes to stable, well-tuned training.
03 Deep Learning Frameworks Implement, track, and tune models in PyTorch, with working knowledge of Keras and JAX.
04 CNNs & Computer Vision Build convolutional networks for classification, detection, and segmentation.
05 Sequence Models & Transformers Model sequential data with RNNs and master the Transformer that powers modern AI.
06 Advanced Architectures & Generative Models Explore modern vision, self-supervised, generative, graph, and reinforcement learning models.
07 Hardware & Distributed Training Understand accelerators and train large models efficiently across many GPUs.
08 Large Language Models & Multimodal AI Pretrain, fine-tune, align, and apply LLMs and multimodal foundation models.
09 Inference Optimization & Production Compress, serve, and operate deep learning models reliably at scale and on the edge.
10 Capstone & Career Demonstrate end-to-end deep learning skills and prepare for engineering interviews.
Deep Learning Engineer
For engineers who want to design, train, scale, and deploy deep neural networks. You will be able to build vision, sequence, generative, and LLM systems and run them efficiently in production.
10 STAGES · 40 TOPICS · 168 SUBTOPICS
1.1 Programming & Software Engineering
Python Programming Data Structures · OOP · Typing · Async
Data Manipulation NumPy · Pandas · Polars
Linux & Command Line
Version Control Git · GitHub · GitLab
Clean Code & Design Patterns
1.2 Applied Mathematics
Linear Algebra Vectors · Matrices · Eigenvalues · SVD
Calculus Derivatives · Partial Derivatives · Chain Rule
Probability & Statistics Distributions · Bayes' Theorem · MLE
Information Theory Entropy · Cross-Entropy · KL Divergence
1.3 Machine Learning Basics
Supervised vs Unsupervised Learning
Linear & Logistic Regression
Classical Models SVM · Decision Trees · k-NN
Ensemble Methods Random Forest · Gradient Boosting · XGBoost
Bias-Variance & Overfitting
Evaluation Metrics Accuracy · Precision · Recall · F1 · ROC-AUC
2.1 Artificial Neural Networks
Perceptrons & Multi-Layer Perceptrons (MLPs)
Activation Functions Sigmoid · Tanh · ReLU · GELU · Swish
Forward Propagation & Backpropagation
Computational Graphs & Autodiff
2.2 Training Neural Networks
Loss Functions MSE · Cross-Entropy · Huber · Focal
Weight Initialization Xavier · He · Orthogonal
Normalization BatchNorm · LayerNorm · GroupNorm · RMSNorm
Regularization L1 · L2 · Dropout · Early Stopping · Weight Decay
Debugging Training Vanishing Gradients · Gradient Clipping · Overfit One Batch
2.3 Optimizers & Schedules
Gradient Descent Variants Batch · Mini-batch · SGD · Momentum · Nesterov
Adaptive Optimizers AdaGrad · RMSprop · Adam · AdamW
Large-Batch & Modern Optimizers LAMB · Lion · Muon
Learning Rate Schedules Warmup · Cosine Annealing · One-Cycle
3.1 PyTorch
Tensors & Autograd
Building nn.Module Classes
Datasets & DataLoaders
Training Loops & Checkpointing
Higher-Level Wrappers PyTorch Lightning · Hugging Face Accelerate
3.2 TensorFlow & Keras
Eager Execution vs Graph Mode (tf.function)
Keras 3 APIs Sequential · Functional · Subclassing
Input Pipelines (tf.data)
3.3 JAX Ecosystem
JAX Transformations jit · grad · vmap · shard_map
Neural Network Libraries Flax NNX · Equinox
Optimization with Optax
3.4 Experiment Tracking & Tuning
Experiment Tracking TensorBoard · Weights & Biases · MLflow
Grid & Random Search
Bayesian Optimization Optuna · Ray Tune
Population Based Training (PBT)
4.1 Convolution Fundamentals
Convolutions, Padding & Strides
Pooling & Receptive Fields
Data Augmentation Flips · Crops · Mixup · CutMix
4.2 CNN Architectures
Classic Architectures LeNet · AlexNet · VGG
Modern Architectures ResNet · Inception · EfficientNet · ConvNeXt
Transfer Learning & Fine-Tuning timm · torchvision
4.3 Object Detection
Two-Stage Detectors R-CNN · Faster R-CNN
One-Stage Detectors YOLO Family · SSD · RetinaNet
Detection Metrics IoU · NMS · mAP
4.4 Image Segmentation
Semantic Segmentation FCN · U-Net · DeepLab
Instance & Panoptic Segmentation Mask R-CNN · Mask2Former
5.1 Recurrent Networks
Recurrent Neural Networks (RNNs)
LSTMs & GRUs
Bidirectional & Stacked RNNs
Seq2Seq with Attention Additive · Multiplicative
5.2 Embeddings & Tokenization
Static Embeddings Word2Vec · GloVe · FastText
Character-Level Embeddings
Contextual Embeddings (ELMo)
Tokenization Algorithms BPE · WordPiece · SentencePiece · tiktoken
5.3 Transformer Architecture
Self-Attention & Multi-Head Attention
Positional Encodings Sinusoidal · Learned · RoPE · ALiBi
Residuals & Feed-Forward Layers
Encoder-Only Models BERT · RoBERTa · DeBERTa · ModernBERT
Decoder-Only Models GPT · Llama · Mistral · Qwen
Encoder-Decoder Models T5 · BART
5.4 State Space Models & Alternatives
State Space Models S4 · Mamba · Mamba-2
Hybrid Architectures (Jamba)
RWKV & Linear Attention
6.1 Advanced Computer Vision
Vision Transformers ViT · Swin · DeiT
Transformer Detectors DETR · RT-DETR
Promptable Segmentation SAM · SAM 2
3D Vision NeRF · Gaussian Splatting · Point Clouds
Video Understanding Action Recognition · VideoMAE
6.2 Self-Supervised Learning
Contrastive Learning SimCLR · MoCo · BYOL
Masked Image Modeling MAE · BEiT
Self-Distillation DINO · DINOv2
6.3 Autoencoders & GANs
Autoencoders Vanilla · Sparse · Denoising
Variational Autoencoders (VAEs)
GAN Fundamentals Generator · Discriminator · Adversarial Loss
GAN Variants DCGAN · WGAN · StyleGAN
Image-to-Image Translation Pix2Pix · CycleGAN · Style Transfer
6.4 Diffusion & Flow Models
Diffusion Fundamentals DDPM · DDIM · Noise Schedules
Classifier-Free Guidance
Latent Diffusion Stable Diffusion · Flux
Diffusion Transformers (DiT)
Flow Matching & Rectified Flow
6.5 Graph Neural Networks
Graph Convolutional Networks (GCNs)
Message Passing Neural Networks (MPNNs)
Graph Attention Networks (GATs)
GNN Libraries (PyTorch Geometric)
6.6 Deep Reinforcement Learning
MDPs & Q-Learning
Deep Q-Networks (DQN)
Policy Gradients & Actor-Critic PPO · SAC
Offline & Inverse RL
7.1 Hardware Accelerators
GPU Fundamentals CUDA · cuDNN · ROCm
GPU Architecture Tensor Cores · HBM · NVLink
TPUs & Custom ASICs Google TPU · AWS Trainium · AWS Inferentia
Inference Chips NPUs · Groq LPU
7.2 GPU Programming & Profiling
CUDA Kernel Basics
Custom Kernels with Triton
Profiling PyTorch Profiler · Nsight Systems
7.3 Efficient Training
Mixed Precision FP16 · BF16 · FP8
Gradient Checkpointing & Accumulation
Efficient Attention (FlashAttention)
Graph Compilation torch.compile · XLA
7.4 Distributed Training
Data Parallelism (DDP)
Model Parallelism Tensor · Pipeline
Sharded Training FSDP · DeepSpeed ZeRO
Context Parallelism (Ring Attention)
Distributed Tooling torchrun · Accelerate · Megatron-LM
8.1 LLM Foundations
Pretraining Objectives Causal LM · Masked LM
Scaling Laws & Emergent Abilities
Mixture of Experts (MoE)
Long Context RoPE Scaling · YaRN
Reasoning Models Test-Time Compute · Verifiable Rewards
8.2 Fine-Tuning & Alignment
Supervised Fine-Tuning (SFT)
Parameter-Efficient Fine-Tuning LoRA · QLoRA · DoRA
RLHF Reward Models · PPO · GRPO
Preference Optimization DPO · ORPO · KTO
Fine-Tuning Tools TRL · PEFT · Unsloth · Axolotl
8.3 Building with LLMs
Prompting & In-Context Learning
Retrieval-Augmented Generation Embeddings · Vector Databases · Reranking
Agentic Workflows ReAct · LangGraph · LlamaIndex · MCP
LLM Evaluation Benchmarks · LLM-as-a-Judge
8.4 Multimodal Models
Contrastive Image-Text Models CLIP · SigLIP
Vision-Language Models LLaVA · Qwen-VL · InternVL
Visual Question Answering (VQA)
Speech & Audio Models Whisper · WaveNet · VITS
Video Generation Sora · Veo
9.1 Model Optimization
Quantization PTQ · QAT · GPTQ · AWQ · GGUF
Pruning Magnitude · Structured · Unstructured
Knowledge Distillation
KV Cache & PagedAttention
Speculative Decoding
9.2 Inference Runtimes & Edge
ONNX & ONNX Runtime
NVIDIA Stack TensorRT · TensorRT-LLM
Intel Hardware (OpenVINO)
Mobile Frameworks LiteRT · Core ML · ExecuTorch
Edge Devices Jetson Orin · Raspberry Pi
9.3 Model Serving
REST APIs FastAPI · Flask
gRPC & Protocol Buffers
Inference Servers Triton · vLLM · SGLang · Ray Serve · BentoML
Dynamic Batching & Autoscaling
9.4 MLOps Infrastructure
Docker & Container Best Practices
Kubernetes for ML GPU Scheduling · KServe
ML Pipelines & Lifecycle Kubeflow · MLflow
CI/CD for Machine Learning
Data Version Control (DVC)
Feature Stores Feast · Hopsworks
9.5 Monitoring & Maintenance
Data Drift & Model Drift Detection
Logging & Telemetry Prometheus · Grafana · OpenTelemetry
Safe Rollouts A/B Tests · Shadow Deployments · Canary Releases
10.1 Portfolio Projects
Architectures from Scratch MLP · CNN · Transformer
End-to-End Vision System
Fine-Tune & Deploy an LLM
Reproduce a Research Paper
Open Source Contributions PyTorch · Hugging Face
10.2 Research Skills
Reading Papers arXiv · Hugging Face Papers
Experiment Design & Ablations
Technical Writing & Blogging
10.3 Job Preparation
Resume & Portfolio
Deep Learning Theory Interviews
Coding Interviews Python · PyTorch · DSA
ML System Design
Take-Home Assignments