Inspiration
Prescription errors in India cause 5.2 million injuries every year, and almost 70% of doctors still write handwritten prescriptions that are misunderstood or misread. We realized that doctors are extremely overworked (India has 1 doctor per 834 people) - so even a few minutes saved per patient can make a massive impact. We wanted to build something that reduces errors, saves time, and makes healthcare safer instantly.
What it does
VOK converts a doctor’s spoken words into a clean, accurate prescription using Deepgram speech-to-text + Gemini medical reasoning. Doctors can: Dictate prescriptions in seconds Auto-generate structured and readable prescriptions Edit anything instantly Share it with patients through a QR code It’s fast, accurate, and removes handwriting errors completely.
How we built it
Used Deepgram’s real-time speech-to-text API to capture doctor dictation with high accuracy. Passed the transcript to Gemini’s medical model for structured prescription formatting. Built a Flutter-based interface where doctors can edit, finalize, and generate a QR code for easy patient access. Designed a clean UI focused on speed and simplicity for doctors.
Challenges we ran into
Getting medical terms recognized perfectly in speech-to-text. Making the output prescription universal, clean, and safe across different accents. Ensuring the workflow is fast enough for real-world use inside clinics. Balancing high accuracy with limited build time.
Accomplishments that we're proud of
Achieved highly accurate medical transcription using a combination of Deepgram + Gemini. Created a usable doctor-first workflow that actually feels faster than writing. Built an end-to-end solution from voice → prescription → QR share in a short timeframe. Designed something that could genuinely reduce prescription errors and help millions.
What we learned
Real-time speech-to-text requires careful tuning for different speaking styles. Healthcare UX is very different - speed and clarity matter more than features. Combining multiple AI systems (Deepgram + Gemini) creates much more reliable results. Problem-solving in healthcare means thinking deeply about accuracy, safety, and trust.
What's next for VOK
Adding multi-language support (Hindi + regional languages). Integrating doctor authentication and patient record storage (encrypted). Adding voice-based follow-up notes and diagnosis suggestions. Testing with real doctors to improve accuracy further. Eventually creating a complete AI-powered assistant for modern clinics.
Log in or sign up for Devpost to join the conversation.