Muhammad Moosa Khan@contactmoosakhan26763
1
Events attended
1
Submissions made
Canada
4+ years of experience
About me
Software Engineering student at the University of Waterloo, building things that actually work — and occasionally things that just look cool. I've done AI research, embedded systems, full-stack web dev, and a brief but distinguished career teaching Python. Currently an AI Engineer Intern @ Presto Phoenix where I can ship real code and break things responsibly. Off the keyboard: you'll find me losing sleep over basketball, getting destroyed at chess (it's called 'learning'), strip-mining Minecraft worlds at 2am, and eating my way through every cuisine known to humanity. If you know a good spot, I'm already on my way.
🤓 Latest Submissions
Skribe — AI-Assisted Medical Intake
Skribe reimagines the medical intake process using conversational AI. Instead of filling out generic forms, patients have a voice conversation with Skribe — a warm, professional AI agent that asks smart, clinically grounded follow-up questions based on NIH MedQuAD data (47,457 medical Q&A pairs). Why I built Skribe: As a first-generation immigrant, I've seen firsthand how language barriers make medical visits overwhelming. Explaining symptoms in a second language is hard — and when patients struggle, doctors miss critical information. Skribe gives every patient a patient, multilingual voice that speaks their language and asks the right questions, so nothing gets lost before the appointment even starts. North America also faces a growing shortage of medical staff. Skribe doesn't replace nurses — it helps them. By handling structured intake automatically, Skribe frees up clinical staff to focus on what only humans can do: care. How it works? Skribe conducts a two-phase session: first collecting patient history (name, age, conditions, medications, allergies), then diving deep into today's symptoms using RAG-retrieved clinical knowledge to ask the right follow-up questions every time. The result is a structured, physician-ready PDF report with patient history, symptoms, severity, triggers, associated symptoms, and a plain-English summary — shareable via a unique link before the appointment. Built with Llama 3.3 70B on Groq, OpenAI Whisper for speech-to-text, ElevenLabs for text-to-speech, ChromaDB for vector search, FastAPI, and React. Why not just use any other LLM? A generic LLM chats about symptoms but has no clinical structure, no memory, and no boundaries — it might even suggest a diagnosis, which is dangerous. Skribe is different: it follows the OLDCARTS clinical framework used by real nurses, grounds every follow-up question in 47,457 NIH medical Q&A pairs via RAG, remembers your medical history across visits but never diagnoses.
10 May 2026
