| 1 |
Fundamentals |
Tensor Basics 101 |
CPU/GPU tensors, dtype conversions |
🌱 Beginner |
30min |
Python basics |
| 2 |
Fundamentals |
Autograd Under the Hood |
Computation graphs, gradient flow |
🌱 Beginner |
45min |
Calculus |
| 3 |
Fundamentals |
Custom Tensor Operations |
torch.autograd.Function |
🔧 Intermediate |
1hr |
OOP in Python |
| 4 |
Computer Vision |
MNIST Classifier in 10 Minutes |
CNN, CrossEntropyLoss, metrics |
🌱 Beginner |
40min |
Neural Networks 101 |
| 5 |
Computer Vision |
Transfer Learning with ResNet-50 |
Fine-tuning, feature extraction |
🔧 Intermediate |
1.5hr |
CNN basics |
| 6 |
Computer Vision |
YOLO-lite Implementation |
Anchor boxes, IoU loss |
🚀 Advanced |
3hr |
Object detection theory |
| 7 |
Computer Vision |
GAN for Anime Face Generation |
Generator/Discriminator, DCGAN |
🚀 Advanced |
4hr |
Probability theory |
| 8 |
NLP |
Text Classification with BiLSTM |
Embeddings, seq2vec |
🔧 Intermediate |
2hr |
RNN basics |
| 9 |
NLP |
Build a BPE Tokenizer |
Subword tokenization, vocabulary building |
🔧 Intermediate |
1.5hr |
NLP preprocessing |
| 10 |
NLP |
Fine-tune BERT for Sentiment Analysis |
HuggingFace integration, CLS tokens |
🚀 Advanced |
2.5hr |
Transformer architecture |
| 11 |
Advanced Architectures |
Graph Neural Networks with PyG |
Message passing, node classification |
🚀 Advanced |
3hr |
Graph theory |
| 12 |
Advanced Architectures |
Neural ODEs for Time Series |
ODE solvers, adjoint method |
🚀 Advanced |
4hr |
Differential equations |
| 13 |
Advanced Architectures |
Implement Transformer from Scratch |
Multi-head attention, positional encoding |
🚀 Advanced |
5hr |
Linear algebra |
| 14 |
Training Optimization |
Mixed Precision Training (FP16) |
Gradient scaling, AMP |
🔧 Intermediate |
1hr |
CUDA basics |
| 15 |
Training Optimization |
Hyperparameter Tuning with Optuna |
Bayesian optimization, pruning |
🔧 Intermediate |
2hr |
Model evaluation |
| 16 |
Training Optimization |
Gradient Accumulation for Large Batches |
Memory-efficient training |
🔧 Intermediate |
45min |
Backpropagation |
| 17 |
Deployment |
ONNX Export for Production |
Model serialization, ONNX Runtime |
🔧 Intermediate |
1.5hr |
Model architecture |
| 18 |
Deployment |
Quantize Model to INT8 |
Post-training quantization |
🚀 Advanced |
2hr |
ONNX knowledge |
| 19 |
Deployment |
FastAPI Model Serving with Async |
REST API, batch processing |
🔧 Intermediate |
2hr |
Web basics |
| 20 |
Reinforcement Learning |
DQN for Atari Pong |
Experience replay, frame stacking |
🚀 Advanced |
5hr |
Q-learning |
| 21 |
Reinforcement Learning |
PPO for Robotic Control |
Policy gradients, continuous actions |
🚀 Advanced |
6hr |
RL basics |
| 22 |
Debugging |
Fix NaN Gradients |
Hooks, tensor inspection |
🔧 Intermediate |
1.5hr |
Autograd |
| 23 |
Debugging |
GPU Memory Profiling |
Memory leaks, caching strategies |
🚀 Advanced |
2hr |
CUDA programming |
| 24 |
Research |
Reproduce AlphaFold Attention |
MSA, pairwise attention |
🚀 Advanced |
8hr |
Bioinformatics |
| 25 |
Research |
Train a Diffusion Model (DDPM) |
Noise schedules, U-Nets |
🚀 Advanced |
6hr |
Probability theory |
| 26 |
Community |
AI-Generated Memes with VQGAN-CLIP |
Text-to-image synthesis |
🔧 Intermediate |
3hr |
GAN basics |
| 27 |
Community |
Real-Time Style Transfer Web App |
OpenCV integration, model serving |
🔧 Intermediate |
4hr |
Flask basics |
| 28 |
Hardware |
Train on TPUs with XLA |
XLA compiler, Google Colab TPUs |
🚀 Advanced |
3hr |
Distributed training |
| 29 |
Hardware |
Deploy on Jetson Nano |
ARM optimization, TensorRT |
🚀 Advanced |
5hr |
Edge computing |
| 30 |
From Scratch |
Build AdamW Optimizer |
Momentum, weight decay |
🚀 Advanced |
2hr |
Optimization math |
| 31 |
From Scratch |
DIY Distributed Training |
All-reduce, NCCL backend |
🚀 Advanced |
4hr |
Multi-GPU basics |
| 32 |
From Scratch |
Micrograd Implementation |
Autograd engine in 200 lines |
🔧 Intermediate |
3hr |
Computational graphs |
| 33 |
Computer Vision |
Semantic Segmentation with U-Net |
Dice loss, patch prediction |
🚀 Advanced |
3.5hr |
Image segmentation |
| 34 |
Computer Vision |
Neural Style Transfer |
Gram matrices, content/style loss |
🔧 Intermediate |
2hr |
CNN feature maps |
| 35 |
NLP |
Named Entity Recognition (BiLSTM-CRF) |
Viterbi decoding, BIO tags |
🚀 Advanced |
4hr |
Sequence labeling |
| 36 |
NLP |
Text Generation with GPT-2 |
Sampling strategies (top-k, temperature) |
🔧 Intermediate |
2hr |
Language models |
| 37 |
Advanced Architectures |
Implement Swin Transformer |
Shifted windows, hierarchical vision |
🚀 Advanced |
6hr |
ViT basics |
| 38 |
Advanced Architectures |
Neural Rendering (NeRF) |
Volume rendering, ray marching |
🚀 Advanced |
8hr |
3D graphics |
| 39 |
Training Optimization |
Prune Models Iteratively |
Magnitude pruning, sparsity |
🔧 Intermediate |
2hr |
Model compression |
| 40 |
Training Optimization |
LR Finder (like fastai) |
Learning rate range test |
🔧 Intermediate |
1hr |
Optimization |
| 41 |
Deployment |
TorchScript for Mobile |
Scripting vs tracing |
🔧 Intermediate |
1.5hr |
Mobile development |
| 42 |
Deployment |
Dockerize XTorch Models |
Containerization, CUDA in Docker |
🔧 Intermediate |
2hr |
Docker basics |
| 43 |
Reinforcement Learning |
World Models with VAE |
Latent dynamics, dreamer architecture |
🚀 Advanced |
7hr |
Variational inference |
| 44 |
Debugging |
Profile Training with PyTorch Profiler |
Flame graphs, bottleneck analysis |
🔧 Intermediate |
1.5hr |
Performance tuning |
| 45 |
Research |
Adversarial Attacks (FGSM/PGD) |
Robustness evaluation, epsilon bounds |
🚀 Advanced |
3hr |
CNN vulnerabilities |
| 46 |
Research |
Quantum ML with PennyLane |
Hybrid quantum-classical models |
🚀 Advanced |
5hr |
Quantum computing |
| 47 |
Community |
Kaggle Pipeline with XTorch |
Custom datasets, submission format |
🔧 Intermediate |
2hr |
Kaggle basics |
| 48 |
Community |
Collaborative Filtering for RecSys |
Matrix factorization, embeddings |
🔧 Intermediate |
3hr |
Recommendation systems |
| 49 |
Hardware |
Benchmark CPU vs GPU vs TPU |
Speed tests, cost analysis |
🔧 Intermediate |
2hr |
Hardware specs |
| 50 |
From Scratch |
Implement LayerNorm Kernel |
CUDA, Triton DSL |
🚀 Advanced |
6hr |
GPU programming |
| 51 |
From Scratch |
Build a Tensor Library (Like NumPy) |
Strides, broadcasting |
🚀 Advanced |
10hr |
Memory layout |