Sunidhi Divekar

I’m an AI/ML engineer with hands-on experience building end-to-end GenAI and deep learning applications. I’ve focused on production-grade retrieval and generation, especially building RAG systems that improve answer reliability through smarter retrieval.

I developed a production-deployed RAG system using FAISS, Gemini Flash, and LangChain, including question-type detection and context reranking. Using chunking plus reranking over FAISS, I reduced hallucinations by ~35–40% versus a naive retrieval baseline, and I designed custom instruction-tuned prompting to improve precision across document types.

Alongside my ML work, I bring strong software engineering discipline: I’ve worked across production Python codebases, wrote modular maintainable automation scripts with clean separation-of-concerns, and collaborated using Git-based workflows and structured code reviews. I’m actively building toward agentic AI systems, and I enjoy shipping practical capabilities with tools like Streamlit, FastAPI, and Docker.