Capture physician dictation and produce structured medical reports without breaking hospital workflows.
Physicians needed to convert spoken dictations directly into structured medical reports without breaking hospital workflows.
Built a Flask backend and Flutter client that capture physician voice recordings and convert them into structured clinical reports used in hospital workflows.
$ render architecture.mmd
flowchart LR Mic[Physician Mic] --> App[Flutter Client] App -- audio chunks --> API[Flask API] API --> ASR[Speech Recognition] ASR --> NLP[Section Classifier] NLP --> Tmpl[Report Templater] Tmpl --> Out[(Structured Report)] Out --> Hospital[Hospital EHR]
$ git log --oneline decisions/
Audio captured and streamed in chunks so transcription started while the physician was still talking — perceived latency near zero.
Outputs follow per-specialty report templates so the EHR import is deterministic and reviewers can scan reports in seconds.
Have a similar challenge?
Let's talk