India · Remote · 6+ years
I design & ship AI agentsthat survive real users — guardrails, observability, and cost control included.
6+
years building software
50+
engineers mentored
60%
faster deployments
35%
fewer production issues
(01) — Positioning
A flashy LLM demo takes a weekend. An AI agent you can actually run in production — that's the real work.
Full-Stack AI Engineer building LLM-powered, agentic systems with LangChain, LangGraph, RAG, and MCP over Claude (AWS Bedrock) and OpenAI. Proven impact in agent guardrails, LLM observability, and cost optimization — shipping well-tested AI systems on AWS and Azure for enterprise clients in fintech, publishing, and telecom.
Selected Work
(02)(03) — How I work
Engineering AI that survives contact with production.
Real-time loop guardrails wrap every tool call — a HookedToolNode plus a HookRunner with a PASS / RESOLVE / REWRITE / OBSERVE model — that catch pathological agent loops and self-correct before they blow the context window.