← Back to Home

ML Systems & Compilers

70-day curriculum — from GPU architecture to TVM, Triton, and distributed training

📅 10 weeks ⏱ 175 hours 🎯 5 phases ⚡ 70 daily lessons

Phases

Phase I

Hardware & Compute Foundations: GPU Architecture → PyTorch Internals Days 1–14

Phase II

Compiler Infrastructure: IRs, Passes, Triton & torch.compile Days 15–28

Phase III

Apache TVM Deep Dive: Relay → TIR → Tuning → MLIR & XLA Days 29–49

Phase IV

Inference Optimization: Quantization, TensorRT & LLM Serving Days 50–63

Phase V

Training at Scale: Distributed Training & Capstone Project Days 64–70

Weekly Schedule