Full-Stack Developer · Gujarat, India
I build AI agents, developer tools and full-stack products —
the kind that ship, not the kind that sit in a slide deck.
Five years of shipping. Four companies. Fifty students trained. One rule — build things people can actually use.
Salient Studio — India · Remote
CodeStorm — India · Remote
Datavizz — Ahmedabad, India
AttainU — Remote, India
Signed, realistic mock Razorpay webhooks fired straight at localhost. Like stripe trigger — but for Razorpay. No ngrok, no dashboard clicking, no real payments. Zero runtime dependencies.
An AI-driven browser control system in Python + Playwright. The model navigates, clicks, reads and extracts data from live websites with no human in the loop — multi-LLM support across OpenAI, Gemini, DeepSeek, Groq and Ollama.
Browser automation breaks the moment a page changes. Selectors rot, flows shift, and every new site means writing the whole script again. Anything built on hard-coded selectors has a shelf life measured in weeks.
Put the model in the driver’s seat instead of the selectors. The agent reads the live page, decides the next action, and executes it through Playwright against real Chrome running the user’s own profile — so existing sessions and logins already work, with no re-auth dance.
It is the difference between a scraper you maintain forever and an agent you point at a goal. The same session that costs pennies on DeepSeek runs free on Gemini’s and Groq’s daily quotas.
A zero-dependency CLI that fires signed, realistic Razorpay webhooks straight at localhost. Published to npm.
Razorpay’s own documentation tells you to expose localhost with ngrok and click real test-mode events in the dashboard. That loop is slow, it is manual, and it cannot run in CI at all — so webhook handling ends up being the least-tested part of a payment integration.
Skip the dashboard entirely. Build the webhook body locally, sign it exactly the way Razorpay signs it, and POST it wherever you point the CLI. Stripe developers have had this for years; Razorpay developers did not.
X-Razorpay-Event-Id — so your existing verification code runs completely unchanged.trigger to fire an event, events to list them, sign and verify to debug a signature by hand.A webhook round-trip that took a dashboard visit and an ngrok tunnel now takes one command and about twelve milliseconds — and it works in CI, which the official workflow never did.
A meme leaderboard where money — not the algorithm — decides who gets seen. Payments, ranking, moderation and the whole product surface.
Every feed is ranked by an algorithm nobody can see. BidYourMeme replaces it with the most honest ranking signal there is: money. Post a meme for free, then outbid the one above you to climb. Want the top spot? Beat the current bid.
The engineering was the easy half. The interesting part was watching what people do when a leaderboard becomes competitive over something completely stupid — which is a product question, not a code one.
A medical assistant chatbot doing symptom analysis and health guidance, wired to the Gemini API with a fast, friendly interface.
A medical assistant chatbot that does symptom analysis and gives general health guidance in plain language, with a fast, friendly interface instead of a wall of disclaimers.
config.js and the whole thing runs at no cost.It is a guidance tool and a demonstration of conversational UX over an LLM — not a substitute for a doctor, and it says so.
Full-stack social platform — authentication, CRUD, profile management and post interactions on a responsive React frontend over secure Express APIs.
Build a social platform end to end — not a UI mock over fake data, but real authentication, real persistence and the interaction loop that makes a feed worth opening.
It is the project where the full stack stopped being four separate skills and started being one.
A complete commerce API on Node, Express and MongoDB — catalogue, cart, orders and auth, structured the way a real store needs it.
A commerce API structured the way a real store needs it, rather than the way a tutorial ends it — catalogue, cart, orders and authentication as one coherent system.
A browser game about the worst day at work, a cyberpunk demo site, dashboards, CLIs and experiments. I build in public — the repo list is the honest version of the résumé.
Systems where a model does the work instead of a brittle script — browser agents, LLM-backed workflows, extraction pipelines that survive a page redesign.
Empty repository to deployed product. React on the front, Node and MongoDB behind it, and the unglamorous middle — auth, roles, payments, tests — actually finished.
The boring infrastructure that makes a team faster. If your workflow has a step everyone dreads, that step is usually a CLI away from disappearing.
I started in a full-stack apprenticeship in 2022, writing React and Node until the tests passed at 95% coverage. Since then I've built dashboards at an agency and spent two years shipping production web products in React, Node and TypeScript.
These days most of my energy goes into AI-adjacent engineering — autonomous browser agents, LLM-backed products, and developer tools that remove the annoying part of someone's day. I now build at Salient Studio, a small AI product engineering studio, and publish the tools I wish existed: razorpay-trigger is on npm.
An MSc in Marketing from Edinburgh Napier sits behind the engineering, which is a strange combination until you watch me name a product, price it and ship it in the same week.
Open to full-stack and AI engineering roles and product work — remote, UK, US, Germany or India. Or just send an idea you've been sitting on.
bhattiujjval007@gmail.com ↗