Team Lead & Senior React / Next.js engineer based in Kanpur, India. 4 years shipping production React at Fluid AI. Open to full-time roles and freelance engagements — say hi if you're hiring or building.
— what I build with —
I'm Tarun. I've been writing React for about four years — long enough to have strong opinions about folder structure, short enough that every new Next.js release still feels exciting.
These days I work on a platform that lets enterprises run their own AI assistants. The kind of app where one bad render costs someone's afternoon, so I've learned to care about caching, re-renders, and the small decisions that compound across a hundred screens.
The web still feels like a playground if you let it.
On the side, I build end-to-end applications — including a RAG Knowledge Assistant, a Tool-Calling Research Agent, and a QR digital dining platform.
A selection of my active applications, from AI-powered research and document search to digital dining.
Ask questions across your own documents
A document question-answering app that accepts pasted text and uploads, extracts and chunks content, stores embeddings, retrieves relevant passages, and streams grounded answers through a separate API.
An agent that searches, reads, and reports
A streaming research assistant that calls web-search and page-extraction tools, then returns findings in a chat interface. The app includes a separate API and persists conversations and messages.
Turn source material into short-form scripts
An AI application that takes user-provided source text and generates hooks, short-form video scripts, and captions through a Next.js interface and an Express API.
A guest ordering flow backed by a restaurant API
A multi-app QR dining project spanning a customer interface, staff dashboard, and NestJS API. The backend models restaurants, menus, dining sessions, and orders, with signed table access and staff authentication.
Four years at one company, three different roles, a lot of late nights.
Owning a big chunk of the Fluid GPT frontend — performance, multitenancy, and the move to on-prem deployments. Most days are equal parts code, code review, and arguing about state-management patterns.
Got the cloud app running fully on-prem so enterprise clients can host it inside their own networks.
Replaced ConfigCat with a custom tenant-level feature-flag system that reads from env vars at boot.
Migrated to RTK Query and watched API calls drop ~85% and page loads improve ~80% on the heavy modules.
Spent way too long getting the Playwright suite to run reliably on tenant-aware fixtures. Worth it.
Four areas I keep returning to. Hover any chip for a bit of context.
Pixels, motion, and the parts users feel.
APIs, sockets, auth — the moving parts.
Code from machine to production.
Habits, not just tools.
Roles, freelance gigs, side-project collabs, or just a hello — I read everything and usually reply within a day.