AI is easy to demo. Harder to deploy. Hardest to trust. I build for the third category.
Over 3+ years, I've designed and shipped production AI systems in EdTech and E-Commerce, not Jupyter notebooks, but systems that handle real users, real failure modes, and real stakes.
What I've built:
- RAG-based study assistants (60% faster content discovery, 15% higher course completion)
- Text-to-SQL conversational analytics (80% reduction in report generation time)
- LLM-based mentor quality auditing (90% less manual QC)
- ECCN Trade Compliance RAG classifier (Recall@5 improved from 65% to 86.5%)
Tech stack I live in: LangChain, LangGraph, CrewAI, FastAPI, FAISS, ChromaDB, Gemini, OpenAI, Python, Docker, AWS.
Currently at Great Learning as an AI Engineer where I'm building AI infrastructure that actually moves business metrics.