Writing & Research
Notes on my experiences, life, and sometimes machine learning.
Writing
- Training Models Across Multiple Machines Building a federated learning system from scratch with gRPC and Protocol Buffers.
- Distributed Data Parallelism Scaling training across multiple GPUs, and what actually changes in the code.
- Fault Detection in Induction Motors A research internship at IIT Roorkee. Signal processing and ML to predict machinery failure.
- Improving Solar Cell Efficiency A water-based cooling mechanism for solar panels, at the Design and Innovation Center.
Nothing here matches that.
Lecture notes
One chapter per lecture. Rough, long, and mostly written for me. Series marked 🔒 open with a passphrase.
- RL Notes Series Reinforcement learning from Bellman equations through to PPO/GRPO alignment for LLMs.
- Deep Generative Models Notes IISc's DGM lectures. The divergence-minimisation recipe behind every generative model, worked through to GANs, VAEs, diffusion and flows.
- Deep Unsupervised Learning (CS294-158) Berkeley's CS294-158. Why learn without labels at all, compression as the definition of understanding, and the map of the generative families.