Full stack engineer. Node.js, Next.js, Laravel, and AI systems built on LLMs, agent tooling and MCP servers.
I have been shipping production software since 2016, across seven teams and for clients on four continents: the United States, the United Kingdom, Europe, Asia, Australia and New Zealand. Greenfield work is fun, but most of that decade has been spent inside systems somebody else started. Feeds that arrive malformed, databases that grew without a plan, sites that got compromised over a weekend.
For the last two years a growing share of that has been AI work: building on frontier models, wiring agent tooling into existing products, and writing MCP servers so those agents can reach the systems that actually hold the data. It has changed how fast I can move through unfamiliar code, which is most of what this job is.
I work across the whole stack because the interesting failures happen between the layers, not inside them.
LLM features, agent tooling and MCP servers wired into real products.
React, Next.js, Angular. Design systems and motion.
Node, NestJS, Express, Laravel, PHP. REST and GraphQL.
Docker, AWS, DigitalOcean, managed Postgres, CI.
Linux hardening, incident response, malware cleanup.
Most of what I have built is still running. That is the part of the job nobody photographs.
The migration that did not lose a row. The integration that survived the vendor changing their API without telling anyone. The site that stayed up the week it went viral. Lately, the agent that reads a codebase properly because someone wrote it a real MCP server.
None of that is visible from the outside, which is why the rest of this page shows you the systems rather than the screenshots.
01
Full stack and infrastructure
A public tool that audits a website and turns the result into a qualified lead. I built the analysis backend, the reporting frontend, and the deployment path that runs it in production and staging.
The audit runs long, so the API answers immediately and scores in the background. The frontend polls rather than holding a connection open.
02
Backend and platform
An ordering platform rebuilt from the previous year's codebase. Split frontend and backend deployment, managed Postgres, object storage for assets, and an end to end test harness that runs against real user roles.
Product media is served straight from object storage to the browser rather than proxied through the API, which keeps request latency off the catalogue pages.
03
Architecture and integration
A headless rebuild of an agency site fed by a property feed that arrives as XML over FTP. The interesting problem was not the frontend, it was making an unreliable upstream look reliable to the people browsing listings.
The feed is the weak link, so nothing reads from it live. A scheduled worker normalises each drop into Postgres, and the site only ever reads a known-good copy.
Ten years, still running.
Start a conversation→Ten years of production systems, and two of building AI into them. Tell me what is breaking.
Start a conversationQuestions
Talha Muneer is a full stack and AI engineer based in Lahore, Pakistan, with ten years of production experience since 2016. He builds and maintains web systems in Node.js, Next.js, Laravel and PHP, and has delivered projects for clients across four continents, including the United States, the United Kingdom, Europe, Asia, Australia and New Zealand.
He designs, builds and rescues production web systems end to end: frontend interfaces, backend APIs and databases, deployment infrastructure, and server security. Most of his work involves existing systems that need to keep running while being changed underneath, rather than greenfield projects.
Yes. Since 2024 a growing share of his work has been AI engineering: building features on frontier large language models, wiring agent tooling into existing products, and writing Model Context Protocol (MCP) servers so agents can reach the systems that hold the real data. This is applied AI inside shipping products, not prototypes.
An MCP server is a standard adapter that lets an AI model call your real systems, such as a database, an internal API or a file store, instead of guessing from its training data. Businesses need one when they want an AI assistant to answer from their own live data with permissions and audit trails intact.
Frontend: React, Next.js and Angular. Backend: Node.js, NestJS, Express, Laravel and PHP, with REST and GraphQL APIs. Data: PostgreSQL, MySQL and MongoDB. Infrastructure: Docker, AWS, DigitalOcean and hardened Linux servers. AI: frontier LLMs, agent tooling and MCP servers.
Yes, for selected client projects and senior engineering roles. He works remotely from Lahore with clients in the United States, the United Kingdom, Europe, Asia, Australia and New Zealand, and is used to working across those time zones. He typically replies to enquiries within one working day, and the contact form on this site reaches him directly.
Yes. Incident response is a regular part of his work: identifying injected scripts and backdoors, cleaning the database and filesystem, closing the entry point, and hardening the Linux server so it does not recur. He has recovered multiple compromised site estates rather than single pages.
Yes, and it is the work he is best at. That includes databases that grew without a plan, integrations with vendors who change their APIs without warning, and migrations that have to complete without losing a row or taking the service offline.
He is based in Lahore, Punjab, Pakistan, and works remotely with clients worldwide. His current role is with Reign Media in Perth, Australia, and he has delivered projects for clients in the United States, the United Kingdom, Europe, Asia, New Zealand and Pakistan.
Ten years, starting in January 2016, across seven engineering teams and clients on four continents. Roles have ranged from PHP developer to MEAN stack developer to senior software engineer, and the work has spanned e-commerce, healthcare scheduling, gaming platforms and lead generation tools.
Available for select client work and senior engineering roles. If you have something already running that needs to keep running, or you want AI actually wired into it rather than bolted on the side, that is the conversation I am best at.
