At Motadata, I integrated Jev (TypeSafe AI) in place of free-form LLM output, cutting AI decision-making costs by 17% and delivering faster responses. I also prototyped a DeepSeek-based agent harness for ServiceOps and led R&D for synthetic API monitoring in ObserveOps.
At RapidOps Inc., I built a production RAG semantic search system with LangChain, Qdrant, and FastAPI, reducing irrelevant results by about 40% at sub-150ms query latency across 30K+ SKUs. I also engineered embedding pipelines and containerized the AI microservices stack with Docker and CI/CD.
In LexAI, I developed an agentic RAG pipeline for Indian legal queries, achieving 1.0 precision and 0.92 F1. My other projects include an auditable financial risk engine and an NLP fake-news classifier.
