Production AI systems that work in the real world. RAG, AI agents, and voice AI for legal, energy, and SaaS teams who need it live, not a demo.
I'm Muhammad Haseeb, an AI engineer based in Islamabad, Pakistan, specializing in production Agentic RAG systems, AI agents, and voice AI infrastructure.
Most AI projects fail after the demo. They work in a notebook, then break the moment real users, real data volume, or real accuracy requirements show up. My work is the opposite of that. I build systems designed to survive contact with production from day one: proper retrieval architecture, evaluation pipelines, observability, and multi-tenant security where it's needed.
A few things I've shipped:
- An AI legal assistant for Saudi regulatory law that hit 97-98% retrieval accuracy and helped close a $600K pre-seed round
- An enterprise search system indexing 300+ GB of petroleum engineering documents, cutting engineer research time by 95%
- A multi-tenant AI SaaS platform with PostgreSQL row-level security and RBAC enforced at the vector-search layer
- A bilingual English/Urdu voice agent on LiveKit, deployed live at an industry exhibition
Work Terms
Availability: Full-time freelance, remote. Based in Islamabad (PKT, UTC+5), with working hours that overlap US and EU business hours for daily standups or calls.
Communication: Async-first through Guru WorkRooms or Slack, with a scheduled call for kickoff and any major milestone. I respond within 24 hours on business days, usually faster.
Payment: Milestone-based for fixed-scope projects, so you're never paying for work you haven't seen. Hourly for ongoing or fractional engagements, tracked and reported through Guru.
Process: Every engagement starts with a short discovery call to confirm scope and data before any code gets written. I don't start building against an unclear spec, since that's where projects go over budget.
If a request falls outside what I can deliver well, I'll say so directly instead of taking it on and figuring it out along the way.