Muhammad Anas

I build AI systems that actually make it into production, not just proof-of-concept demos that look good in a demo video and fall apart the moment someone relies on them. My main focus is designing multi-agent setups where different LLMs work together instead of leaning on one model to do everything. I did this for a radiology report pipeline where three models run in parallel to cross-check findings before a synthesis agent puts together the final output, with patient data redaction built in from the start. I took the same approach with automating research paper writing, where separate agents handle the literature review, outlining, and section drafting so the whole thing isn't one long manual slog.
Beyond the AI side, I'm a genuine full-stack developer. I can take a product from a rough idea to something people actually use, handling everything from the interface (React, Next.js, Flutter) down to the backend and database work (FastAPI, Supabase, PostgreSQL, MongoDB). THEETIFY, my EdTech app, is a good example of this coming together: an AI-powered study tool with adaptive difficulty and spaced repetition, built as a real working product rather than a class assignment.
I'm also comfortable with the more practical, less glamorous side of automation. I previously automated a company's hiring pipeline, which cut down a significant chunk of manual work in sourcing and screening candidates. So if you need a workflow built around n8n, or a script that saves hours of repetitive effort, that's well within what I can do too.
What I think makes me a good fit isn't just knowing the tools. It's that I tend to think about the parts people forget, like what happens when the AI gets something wrong, or how the system should degrade gracefully instead of just breaking. That habit shows up across my projects, whether it's a safety check in a medical pipeline or clean error handling in a web app.