Projects that move from idea to system.
A growing archive of forecasting pipelines, computer vision products, agentic workflows, and the experiments that make the next project better.
The main thread.
These are the projects that best show how I think: start with the real constraint, build the smallest useful system, and document what still needs work.
Decoder-Only Transformer: MHA vs GQA vs MQA
A GPT-style decoder-only Transformer built from scratch in PyTorch to study the trade-off between language-model quality and autoregressive inference efficiency.
Day-ahead electricity load forecasting
A probabilistic forecasting system designed around the realities of a day-ahead horizon, not just a leaderboard score.
Serengeti classifier
A full data-to-demo computer vision system with honest splits, explainable predictions, and a usable interface.
Rossmann sales
A leakage-safe tabular forecasting pipeline built around temporal features, careful handoffs, and reproducible inference.