Medical Viva Simulator
AI-powered medical examination simulator that provides realistic viva voce practice with real-time grading and feedback for medical students.
Client: Surgicalminds (surgicalminds.com) Industry: Medical Education / Healthcare
The Challenge
Viva voce exams are among the hardest parts of surgical training to prepare for. Unlike written papers, a viva tests how a candidate thinks out loud under pressure — clinical reasoning, structured communication, and the ability to defend an answer to an examiner in real time. For trainees sitting fellowship exams like the FRCS viva, the only realistic way to practise has traditionally been with a senior surgeon sitting across the table, which is expensive, hard to schedule, and impossible to do on demand the night before an exam.
That leaves most candidates under-rehearsed in the one skill the exam actually measures: speaking. They know the material; they just haven't practised performing it.
The Solution
We built an AI-powered viva simulator for Surgicalminds that lets a trainee practise out loud, any time, with an examiner that responds like the real thing.
What it does
- A candidate speaks to an AI examiner in real time — no typing, just conversation, the way the exam runs.
- Live speech recognition transcribes the candidate's answers as they talk.
- The AI examiner asks follow-up questions, probes weak answers, and adapts to the candidate's responses, mirroring how a real panel pushes on reasoning.
- At the end of each session it grades the performance and gives structured feedback across clinical reasoning, communication, and medical knowledge — pinpointing exactly where an answer fell short.
- Sessions can be repeated unlimited times, on demand, so practice is no longer bottlenecked by examiner availability.
How we built it
Real-time voice was the hard part. We used WebRTC for low-latency audio capture, paired with speech recognition to transcribe in real time and OpenAI models to drive the examiner's questioning and grade each response against exam criteria. A Node.js backend orchestrates the session, scoring, and feedback, with a React front end that keeps the experience fast and conversational rather than clunky and turn-based.
Results
- Now used by 900+ medical students and surgical trainees to rehearse their vivas on demand.
- Practice no longer depends on a senior surgeon's calendar — candidates can run a realistic mock viva whenever they need one.
- Every session ends with structured, criteria-based feedback, so candidates know precisely what to fix before exam day.
- Unlimited, repeatable practice drops the marginal cost of one more mock viva to near zero.
This project is part of how we build AI agents and custom software for real-world workflows. If you're exploring something similar, book a free consultation.