From recommendation systems at scale to production LLM agents — a full-stack view of what it takes to build AI products that work in the real world.
Recommendation systems, NLP modeling, and A/B experimentation at scale. Learned what production ML actually requires.
Production LLM systems — multi-agent RAG, async inference at 40K users, full LLMOps lifecycle from prompts to blue-green deploys.
End-to-end multi-agent RAG system serving 40,000+ UPSC Civil Services students in production — designed, built, and owned solo.
Hybrid recommendation system — ALS collaborative filtering, TF-IDF content-based NLP, and a LightGBM ranker — personalising the student homepage for 40,000+ UPSC aspirants. Five A/B experiments. +34% CTR lift. The production ML foundation built before the AI engineering work.