Hitesh U.
Agentic AI Developer | LLM Apps, RAG, Multi-Agent Systems
I build AI agents that do the work — LLMs that plan, retrieve, call tools, and complete multi-step tasks reliably enough for real users. Not demos. Production systems. 15+ years shipping backends in fintech, e-commerce, fleet, and ride-sharing. The last two years focused on the agentic layer. That combination is the point: most AI prototypes break on contact with real traffic, real data, and real auth. Mine don't, because I've shipped the boring half of software for a decade. What I do best Turn manual workflows into reliable agents. Fix RAG that retrieves the wrong thing. Wire LLMs into existing systems — your database, queue, auth, observability — without creating a parallel stack nobody owns. Selected work Multi-agent workflow automation for a fintech platform — replaced a 6-step manual review process, cut turnaround from 2 days to 15 minutes RAG system over 200K+ internal documents with hybrid search and evals — answer accuracy moved from 61% to 92% Custom MCP servers connecting Claude to internal tooling for an e-commerce ops team Spring Boot microservices powering a ride-sharing platform at production scale Stack LangGraph, LangChain, LlamaIndex, CrewAI · Claude, GPT-4/5, Llama 3 RAG, MCP servers, tool use, structured outputs, evals Vector DBs: Pinecone, Weaviate, pgvector, Milvus Java / Spring Boot · Python / FastAPI · Kafka · AWS · Docker · Kubernetes Background 35+ projects delivered, 5-star feedback. B.E. in Computer Engineering. Work daily in Claude Code and Cursor — agentic tooling is how I build, not just what I build. How I work I'll push back if your agent will hallucinate, your retrieval will miss, or your tool schemas will confuse the model — before writing code, not after. I think about latency, cost, retries, and what happens when the model provider goes down. The happy path is the easy half.
Earnings and projects
An estimate from an earlier read
Billed in all: $107,177$107k
Across 57 contracts
Earnings, last 30 days
No data since Nov 5, 2026.