Sanjai Kumar

At NSIC, I built a Python interview assistant with modules for technical explanations, question generation, and interview-answer evaluation. It returns structured feedback with scores, strengths, and suggestions.

I engineered the AI Financial Research Assistant, a RAG pipeline that searches 50 financial reports using embeddings and ChromaDB. On the project test set, it reached 86% retrieval accuracy and reduced hallucinated responses by 82%.

I also developed RAG Guardian, a Python framework for testing LLM applications against prompt injection, unauthorized access, unsupported responses, and valid-answer scenarios. With defenses enabled, it passed all four tests and generated reports with evidence and severity for each result.

I work with Python, LangChain, LangGraph, and FastAPI to build and evaluate LLM applications. I’m seeking a GenAI or Agentic AI Engineer role focused on reliable, production-ready LLM systems.