MediNote AI
A medical-documentation workflow prototype.
Problem and contribution
MediNote AI explored turning a spoken interaction into a structured document. I contributed to this team project combining speech recognition, language-model processing, and report generation. Its intended setting was medical documentation, where preserving the meaning of a conversation is central to the task.
Recorded workflow
The historical project describes capturing speech, transcribing it, processing the transcript with GPT-4, and generating a PDF report. The application account also describes multilingual support and a documentation interface.
Portfolio materials include interface images and a source-code link. They support describing a functional prototype and its components, but do not establish that the generated documents are medically accurate or suitable for autonomous use.
Evidence and limitations
A transcription or language-model error can propagate into a structured report. The materials do not provide a clinical validation study, a documented error rate, or an independently audited privacy and deployment assessment.
This is an earlier documentation prototype. It does not establish clinical deployment, reliable prescription generation, or a replacement for professional review. Dates are omitted because historical sources disagree.