At Deloitte, I developed a real-time agent assist system that processes live phone calls and generates guidance and summaries with Google Gemini, reducing average handling time by ~20%.
I also built and enhanced a multi-agent, MCP-based architecture with FastMCP, connecting enterprise APIs and knowledge bases for dynamic tool calling and contextual retrieval. I designed LLM evaluation frameworks to assess grounding, relevance, and hallucinations across transcripts, APIs, and documents.
Previously, at Tata Consultancy Services, I designed and deployed RAG pipelines and multi-agent systems using OpenAI GPT, LangChain, LangGraph, and LlamaIndex, improving retrieval accuracy by 30%. I also developed ML and deep learning pipelines and deployed model-serving APIs on Azure.
