Starting in the ward, not with the model
The problem was not simply speech recognition. Clinicians were balancing patient care with paper records, noisy rooms, code-switched language, and infrastructure that could not be assumed to stay perfect.
As one of two engineers, I took an AI-powered React Native app from prototype to production, turning doctors’ dictated consultations into patient records. I built the TypeScript APIs and GCP infrastructure, maintaining 99.9% availability over seven months.Built a voice-first hospital app for Urdu and Punjabi, deployed with doctors for seven months.
The problem was not simply speech recognition. Clinicians were balancing patient care with paper records, noisy rooms, code-switched language, and infrastructure that could not be assumed to stay perfect.
I fine-tuned Whisper with LoRA on Urdu and Punjabi clinical audio, then paired transcription with structured extraction so speech could become a usable medical record rather than another block of text.
The same record could do more than archive a visit. I built a retrieval service that matched symptoms against indexed clinical guidance and reranked the evidence before presenting a possible risk to a clinician.
I built the app’s TypeScript REST APIs on Bun and Hono over PostgreSQL, using Drizzle ORM for queries and migrations, Zod for validated data models and OpenAPI for documentation. Pulumi provisioned Cloud Run and Cloud SQL. Each release received traffic only after passing health checks.