I build AI agents and MCP servers, and the production Python they run on.
Most of my work is backend systems for LLM applications — tool-calling agents, Model Context Protocol servers, and the async infrastructure underneath them: FastAPI, SQLAlchemy 2.x async, Celery, Redis, Postgres. I also ship TypeScript front-ends, including one deployed to the Cloudflare Workers edge runtime.
Based in Quetta, Pakistan. Open to contract work on agent and MCP integrations.
netdiag-mcp · Python, MCP SDK 2.x An MCP server giving agents read-only DNS, TLS, HTTP and RDAP diagnostics for any host. Keyless — no API keys to run it. Includes an SSRF guard that refuses non-public targets and re-validates every redirect hop, because these tools take a hostname from a language model and then make a request to it. 65 tests, with the security-critical ones deliberately offline.
Production LLM agent · Python — client work, private A tool-calling agent for a retail platform, built and maintained solo: FastAPI, async SQLAlchemy, Celery workers, Redis, Postgres, Docker. Also an MCP server exposing the same domain to LLM clients.
whatismytools.com · TypeScript, Next.js 16, Cloudflare Workers 18 browser-based network and device diagnostics in 4 languages, deployed to the edge via OpenNext. Everything runs client-side — the interesting constraint is that edge runtimes have no Node APIs, so DNS, geolocation and throughput measurement all had to be rebuilt on web primitives.
Backend — Python · FastAPI · SQLAlchemy 2.x (async) · Pydantic v2 · Celery · Redis · PostgreSQL · Alembic · Docker AI/agents — Model Context Protocol · tool calling · LLM application architecture Frontend — TypeScript · React 19 · Next.js 16 · Tailwind Infra — Cloudflare Workers · Linux · GitHub Actions
Full Stack Developer, National Centre of Robotics and Automation — 2023–present Software Engineer, Government Innovation Lab (UNDP) — 2020–2022
BS Computer Science, BUITEMS (2016–2020) · Winner, Google Android Developers Challenge (2021)
LinkedIn · Stack Overflow · Twitter