Based in India

I build systems
that think.

I build multi-agent LLM applications and automation systems that solve real enterprise problems. My work spans RAG pipelines, agentic workflows, and LLM orchestration — backed by four years of production automation experience at Citibank.

See my work Resume ↓

Open to full-time roles · On-site, Hybrid, or Remote

What I work on

LLM Applications

RAG pipelines, prompt engineering, and agentic workflows using LangChain, LangGraph, Amazon Bedrock, and IBM Watsonx.

AI-Powered Automation

Applying AI reasoning to enterprise automation problems — not just rules and triggers, but agents that understand context. The 66% cycle-time result at Citibank showed what's possible with smarter system design.

Multi-Agent Systems

Designing agents that reason, use tools, and coordinate — not just chat. ReAct-style orchestration with real, structured outputs.

A bit
about me

India
→ wherever the work is interesting

I build AI systems that work in the real world — multi-agent pipelines, RAG applications, and agentic workflows with structured outputs. Alongside that, I have four years of production experience automating enterprise compliance systems for Citibank at Wipro, where one integration I designed cut cycle time by 66% and became the standard approach for the client team.

That depth in how real business processes actually fail — and what it takes to make them reliable — is what I bring to AI engineering work.

I'm looking for a team building serious AI systems. If that sounds like something you're working on, I'd love to talk.

Recent projects

All work →

Personal Project

LoanIQ — Intelligent Loan Approval Assistant

Four specialised AI agents — profiling, risk analysis, decision, compliance — work in sequence to evaluate loan applications and return a structured verdict with risk score, confidence level, and a case ID. Built on LangGraph with Claude as the reasoning engine and MCP servers per agent.

LangGraph Claude (Anthropic) FastMCP FastAPI Streamlit Python
GitHub

AWS-Wipro Codeathon

Revenue Leakage Finder

An agentic system that hunts billing anomalies in telecom data — unbilled usage, pricing mismatches, failed payment cycles. Designed the architecture and built it using Amazon Kiro with a multi-agent setup on Bedrock.

LangGraph Amazon Bedrock Python FastAPI Docker
GitHub

Personal Project

Food Recommendation System

Ask it "what's similar to this?" or "what should I eat for more protein?" — it actually understands. Built on vector embeddings for semantic search and a RAG pipeline for natural language Q&A over nutritional data.

LangChain ChromaDB FAISS Gradio Python
GitHub