Ajitava Das

I’m an Applied AI Engineer focused on building production LLM systems that ship—multi-stage agentic pipelines, RAG, and LLM-as-Judge evaluation. I’ve owned end-to-end architecture, quality gates, and observability for internal business use.

At Sutra.AI, I led development of an “autonomous opportunity-report pipeline” with a 12-stage design using Claude Sonnet + Gemini Deep Research, with grounded outputs and active internal use. I built a 10-criteria quality rubric with automated PASS/CONDITIONAL/FAIL gates to prevent reward hacking, plus a weekly LLM-as-Judge calibration loop to track drift.

I also ship full-stack AI: FastAPI/FastMCP backends, dashboards for feedback capture and calibration tracking, and Langfuse tracing for per-span cost and latency observability. Earlier, I built a Decarbonization Platform with a Next.js dashboard, Climatiq-based emission estimation, and an AI Copilot with dynamic SQL across three LLM providers (with automatic fallback), plus RAG-based document search using LlamaIndex.