I started in PHP in 2021, writing backend modules for client projects nobody would call glamorous. CMS features, e-commerce plumbing, database queries that had grown slow because nothing had ever been measured.
What I noticed early is that the interesting problem is almost never the code. It is that a person somewhere is doing something a computer should be doing, and everyone has stopped noticing because that is simply how it has always worked. Somebody exports a spreadsheet every Monday. Somebody re-types an address into a courier portal. Somebody keeps a private tally because the two systems disagree and they have learned not to trust either.
That became the work I chose. Over five years I moved from backend modules to owning whole systems: the APIs, the queues, the schedulers, the sync engines, and eventually the infrastructure they run on. At Maxenius I led the team that built a catalogue sync holding 300,000 listings in agreement across two storefronts, then led its migration from Laravel to Python.
The last two years pushed me into AI engineering, and it turned out to be the same job wearing different clothes. At Codiux I built and deployed a production voice sales agent on Twilio, Retell AI and OpenAI: real outbound calls, real customers, real consequences when a provider times out mid sentence. Since then I have been building MCP servers so agents can reach internal systems safely instead of through hand written glue.
I build with Claude Code, Codex and Cursor in the loop, and I would rather say so than pretend otherwise. They are very good at the mechanical half of the job. They are not a substitute for knowing why a queue is backing up at three in the morning, and everything that reaches a client is code I have read and can defend.
Calling a model is the easy part. Everything that makes it trustworthy, the latency budgets, the fallbacks, the guardrails on what an agent may actually do, the write backs that make its output useful, is ordinary systems engineering. That is the part I am good at, and it is why I describe myself as a full stack engineer who does AI rather than the other way round.
A system people talk about is usually a system misbehaving. Success is nobody mentioning it for six months.
Anything that can fail will. The question is whether it lands in a queue somebody reviews, or disappears until a customer complains.
Two systems that both believe they are authoritative will eventually disagree. Decide which one wins before writing any code.
If running something twice is safe, retries become free and most distributed systems pain quietly disappears.
The model is one component. Timeouts, retries, permissions and observability decide whether it is usable in production.
Some processes are too rare or too cheap to automate. Saying so costs one project and earns the next three.
Codiux · Lahore
Where the AI work started. Full stack product engineering across Python, React and Next.js, with end to end ownership through to AWS deployment.
Maxenius Solutions · Faisalabad
The engagement that shaped how I work. Large scale commerce integration, the operational automation around it, and my first time leading other engineers.
XLogic Solution · Faisalabad
First role where I owned production services rather than features inside someone else's.
University of Agriculture, Faisalabad · Faisalabad
Software engineering, web development, database systems, computer networks, data structures and algorithms, object oriented programming.
OpenAI, Retell AI, LLM integration
Conversational agents, prompt engineering
MCP servers, built and deployed
Claude Code, Codex, Cursor
Daily, for the mechanical work
Review and architecture stay mine
Python, JavaScript, TypeScript, PHP, SQL
FastAPI, Django, Flask, Laravel
React, Next.js, Tailwind CSS
REST API design, webhooks, microservices
Queue workers, retry and dead letter handling
Twilio, Shopify, eBay, Royal Mail
PostgreSQL, MySQL, MariaDB
MongoDB, Redis, Supabase
AWS: EC2, ECS, RDS, ElastiCache,
Secrets Manager, VPC
Docker, Nginx, Linux VPS, Git
My last engagement ended in August 2026, so I have capacity now for both.
Freelance work: AI agent and MCP builds, integration and automation projects where the outcome is measurable. Hours removed, errors eliminated, a process retired. Those are the projects I do best and enjoy most.
Full time: full stack, backend or AI engineering, remote or onsite in Lahore. I am most useful on teams shipping AI features that have to hold up under real traffic, or where data has to move between systems reliably and somebody needs to own that end to end rather than treat it as glue nobody wants.
Either way, what I care about is whether the work removes real friction for real people. That is a low bar to state and a surprisingly high one to clear consistently.
Either is a good reason to get in touch. Twenty minutes, no pitch.