At XYMA Analytics, IIT Madras, I architect production RAG systems for NPCI/RBI regulatory and industrial corpora, combining dense and sparse retrieval with reranking to improve context precision and reduce hallucination.
I built an Autonomous Financial Dispute Decision Engine using LangGraph state machines, confidence-gated self-reflection, and Pydantic-enforced JSON outputs through Groq and Llama 3.3 70B. I also delivered a UPI Regulatory Dispute Assistant for grounded querying across RBI and NPCI datasets.
I design asynchronous ingestion pipelines for PDF, HTML, and XML documents, embedding enterprise knowledge into ChromaDB and Qdrant. I instrument LLM services with Langfuse and MongoDB Atlas audit logging for latency, token-cost, retrieval-quality, and compliance traceability.
Previously, I built full-stack and AI knowledge-platform applications using Python, Node.js, React.js, TypeScript, MongoDB, and PostgreSQL. I deploy containerized inference services on AWS and GCP using Docker, GitHub Actions, and CI/CD.
