Backend Engineer · Production AI Systems

Backend systems with LLMs in them.
Built to behave in production.

I build backend services and the AI features that run on top of them — focused on latency, reliability, and keeping production uneventful.

Things I've Built

Backend and real-time systems focused on performance and reliability.

Recall

Chat-with-your-documents platform. Upload PDF or DOCX and query them through a citation-backed agent, streamed token by token.

FastAPILangGraphQdrantMySQLMinIOCeleryRedisNext.js

Curie

Live audience-experiment platform, run with 24 participants on their phones.

PythonFastAPIWebSocketsasyncioNumPySciPy

CodePair

Real-time collaborative coding platform over WebSockets.

Real-Time Location Tracker

Live location tracking with instant positional updates.

Work Experience

A timeline of my professional journey and backend experience.

Graduate Engineer — Backend Engineering

OneAssist Consumer Solutions

Nov 2025 — Present · Gurugram, Haryana

Java Backend Developer Intern

OneAssist Consumer Solutions

May 2025 — Nov 2025 · Gurugram, Haryana

Backend Developer Intern

Airtel Digital

May 2024 — Jul 2024 · Gurugram, Haryana

Technical Stack

The technologies I use to design scalable backend systems, integrate intelligent workflows, and maintain reliable production services.

Backend Engineering

Spring BootSpring AISpring SecurityFastAPICeleryRESTMicroservicesSSEWebSocketsSocket.IOExpress.js

AI & LLM Systems

RAGAI AgentsLangChainLangGraphReAct AgentsAzure OpenAIGPT-4oEmbeddingsCross-encoder RerankingHybrid SearchPrompt Engineering

Databases & Storage

MySQLPostgreSQLMongoDBRedisQdrantChromaDBMinIO/S3

DevOps & Tooling

DockerGitHub ActionsJenkinsCI/CDGitGitHubGitLabPostmanJIRAKibana

Languages

JavaPythonSQLJavaScript/TypeScriptC/C++

:?

I work on backend systems — the part of software that users don’t see, but blame anyway.

Right now I'm the sole backend engineer on a benefits platform used across 40+ enterprise clients, which means I own the Spring Boot services behind it, the AI features built on top of them, and the pager when either misbehaves.

Most of what I find interesting sits where those two meet: making a RAG assistant answer in five seconds instead of thirty, deciding what a retrieval system should do when it isn't confident, working out why a service quietly exhausts its heap every other day. Problems where scale, reliability and imperfect inputs collide.

Clean architecture excites me. Debugging production at 1 AM — slightly less.