mohan vishe

At FutureSmart AI, I build production RAG systems, AI agents, and NL2SQL products for enterprise clients. I’ve taken a natural-language database agent and a document-intelligence product through public launch.

I own retrieval, prompting, and evaluation workflows, including benchmark harnesses that track commit, prompt version, and model provenance to prevent regressions from shipping. I also benchmark providers for cost, latency, and accuracy.

I build GenAI APIs with Python and FastAPI, semantic search, Redis caching, document parsing, sentiment analysis, vision features, and speech tools. My work spans Pinecone, Qdrant, Chroma, LangChain, LangGraph, and schema-aware SQL agents.