From manual code to autonomous systems
I'm Muhammad Hamd, also known as Hamd Ali, an agentic AI engineer and systems builder based in Karachi, Pakistan. I design and ship AI-native systems: autonomous agents, LLM integrations, and workflow automation that replace repetitive digital work with software that runs on its own. This is the longer version of how I got here and what I care about building.
It started with backend engineering
My path didn't begin with AI. It began with backend systems. As a Node.js developer at VativeApps (2023–2024), I built scalable Express REST APIs, real-time WebSocket servers, and message queues with RabbitMQ, and integrated third-party services like in-app billing and early OpenAI features into mobile products. That period taught me the unglamorous discipline that still defines my work: writing reliable, maintainable systems that hold up under real traffic, not just code that demos well.
The turn toward agentic AI
At Cubitrek I progressed from full-stack developer to agentic AI engineer over a single intense year. I went from building MERN applications with embedded machine-learning logic to designing autonomous agentic workflows in Python: systems that could plan, orchestrate multiple large language models, and execute multi-step business tasks without a human in the loop. I researched the Google Ads API for data enrichment, managed Azure infrastructure, and spent a lot of time on the hard part of LLM work: making non-deterministic models behave predictably enough to trust in production.
That experience convinced me that the interesting frontier wasn't another chatbot. It was agentic systems: AI that uses tools, keeps memory, reasons across steps, and actually completes work. So I went deeper into the stack that makes it possible: RAG, vector embeddings, orchestration frameworks, and the engineering around them.
Building products, not just features
I'm a builder, so I started shipping my own products. I founded WatBot, a WhatsApp AI automation platform for customer support. Its core engine is written in Go using the whatsmeow library for direct WhatsApp Web integration, with an OpenAI-powered conversational layer and a React dashboard, distributed as a local-first binary. I founded selfbrand AI, a SaaS that automates roughly 80% of personal-branding content for founders and professionals using LLMs. And I'm building Asmara.AI, an AI-native automation product focused on LLM orchestration for business workflows.
Founding products changes how you engineer. You feel every shortcut, every flaky pipeline, every cost spike. It pushed me to care about fallbacks, monitoring, and cost control as first-class concerns, which are the things that separate an AI prototype from a system a business can actually depend on.
Enterprise AI at MindKeepr
Today I'm a full-stack AI engineer at MindKeepr in Tallinn, Estonia (hybrid), where I lead development of AI-powered knowledge-retention systems. I architect agentic AI pipelines with RAG that surface the right information at the right moment, reducing onboarding time and preventing institutional knowledge from being lost when people leave. Working at enterprise scale keeps me honest about reliability, predictability, and safety in mission-critical environments.
What I believe about building with AI
I have a simple philosophy: AI should remove manual work, not add complexity. I'm not interested in hype-driven demos. I'm interested in systems that quietly do real work, reliably, while you sleep. Every technical decision I make is judged against one question: does this deliver practical, measurable utility? Being based in Karachi and working with founders and teams globally, I've learned that the best AI work is grounded in real problems and shipped end-to-end, from architecture to the unglamorous production hardening that makes it last.