I build production frontends in React, Next.js and TypeScript — and the AI and MCP tooling behind them. 7+ years shipping web applications, from data modelling through to the deployed interface.
What I Do
React, Next.js and TypeScript interfaces that work on every screen, covered by unit and Playwright tests, shipped to production.
MCP servers and LLM features built into a product where they earn their place — not bolted on for their own sake.
REST and GraphQL APIs, auth, and data models built to hold up under real traffic.
Scalable architecture from database schema to microservices and event-driven systems.
Background jobs, webhooks and third-party integrations that take the manual steps out of a workflow.
AI integration strategy, code review, architecture audits
Selected Work
A solo-built content-publishing autopilot — draft once and post everywhere (LinkedIn, Threads, Instagram, YouTube) in your own voice, driven entirely from Claude Code in the terminal.
Tech Stack
A multi-tenant SaaS that runs an Indian physiotherapy clinic end to end — patient files, attendance, session packages, GST billing, prescriptions and a self-serve patient portal — from one calm, phone-first screen.
Tech Stack
A remote MCP server that turns Claude into a personal assistant — tracks habits, tasks, finances, goals, documents, and memory with semantic search. Built on OAuth 2.0 + PKCE, deployed serverless on Vercel with 41 MCP tools across 6 domains.
Tech Stack
About
Amar Gupta
Senior Frontend Developer — AI & MCP · Delhi / Remote
I build production frontends in React, Next.js and TypeScript — and the AI and MCP tooling behind them. 7+ years shipping web applications, from data modelling through to the deployed interface.
I focus on system design, scalability, and clean architecture. I use AI tooling like Cursor and Claude Code to move faster, but the fundamentals — data modelling, error handling and tests — are what make a product hold up in production.
Career
Agusta Software Pvt. Ltd
Feb 2023 — Jul 2026 · Pune, India
Reworked a 124-route React 18 loan-servicing portal for phones, and built a Playwright sweep that measures overflow on every route at 375 px: 51 pages went from 287 px of overflow to 0. Wrote 673 Jest and React Testing Library cases, and designed MCP servers that let AI agents reach external tools, databases and APIs.
Spark Eighteen Lifestyle Pvt. Ltd.
Dec 2021 — Aug 2022 · Delhi, India
Worked on a cloud-based project management platform enabling task allocation, team collaboration, and project monitoring. Integrated Express-based APIs with frontend Vue and React apps.
LoudCloud Systems Private Limited
Dec 2020 — Oct 2021 · Mumbai, India
Contributed to a service-based e-commerce platform for a fashion and lifestyle brand, focusing on personalized recommendations and seamless shopping experiences.
Les Transformations Learning Pvt. Ltd.
Sep 2020 — Nov 2020 · Gurgaon, India
Developed an online LMS for a leading EdTech company to provide interactive educational experiences. Built PWA version of LMS features.
Expertise
MCP Server Development, LLM Integration (Claude, GPT), AI Chatbot Building, Prompt Engineering, RAG Pipelines, LangChain, AI Workflow Automation
Microservices Architecture, API Design, Database Schema Design, Scalability Patterns, Event-Driven Systems
React 19, Next.js 16, Vue 3, Nuxt.js, Remix, TypeScript, Tailwind v4, Radix UI, ShadCN
Node.js, Express.js, Supabase, MongoDB, REST APIs, GraphQL, JWT Auth
Redux, Vuex, Pinia, Redux-Saga, Thunk, Zustand
GitHub, Vercel, Netlify, Heroku, Docker, ESLint, Cursor AI, Codex
Shopify Customization, Google Apps Script, PWA, Pine Script, Python (Backtesting)
WRITING
Long-form notes on shipping AI products, MCP servers, and the weird edges of modern web dev.
November 10, 2026 · 5 min read
Three Git projects trended together this week — worktrunk, OmniDiff, and OpenAI's Pando — and they're the same story at three scales: the tooling around Git was never built for the volume of branches, commits and diffs AI coding agents produce, and this week it started catching up.
November 7, 2026 · 5 min read
A frontier model, GPT-6 Astra, was shown reading and rewriting a smaller model's raw weights as text — editing 0.016% of Qwen3.5-0.8B lifted its MATH-500 score from 7.9% to 48.1%. Then the rebuttals and a laptop replication arrived within two days. What the paper actually did, and why reading the primary source first still matters.
November 5, 2026 · 5 min read
GPT-6 Astra claimed to edit a small Qwen model's raw weights by hand — touching 0.016% of them — and send its math scores soaring. A reproduction on a laptop and a sharp critique landed a day later. Here's what was actually measured, what's marketing, and the one lesson that transfers to anyone building with AI.
Get in Touch