I write about ML architecture, production systems, and the decisions that don't make it into the README. Published on Towards Data Science and Hashnode. More coming.
A walkthrough of the full architecture behind PLRS — from the SAKT model and knowledge graph constraint layer through to the FastAPI backend, rate limiting, and CI/CD pipeline. Covers the decisions that mattered and the ones that didn't.
Why prerequisite graphs should enforce ordering in learning systems — not just inform it. Explores how DAG-based constraint layers change the character of recommendations and what the numbers actually look like when you measure it.
What nobody tells you about moving from a RAG prototype to something that runs in production. Chunking strategies, retrieval evaluation, reranking, and the failure modes that only show up at scale.
Diagnosing data leakage in competition features — and what the diagnostic process looks like when you take it seriously. A case study in the difference between EDA and understanding your data.