Skip to main content
AI Automation & Agents· Mobile App

Doctor Notes

The AI pipeline behind a medical-visit recorder: one consultation recording becomes an accurate transcript, a plain-language visit summary, and clinician formats such as SOAP notes and referral letters.

  • Sector Digital health, Medical documentation
  • Client Digital-health company
Becomes a transcript, summary, SOAP note and referral letter
1 recording
Transcription models compared before one was chosen
Evaluated
Diagnoses, tests, medications and next steps patients can follow
Readable
Doctor Notes medical documentation: consultation recording to transcript, plain-language summary and clinical SOAP notes
The starting point

The challenge

Patients forget much of what is said in a consultation, and clinicians lose hours to notes and letters. Both need accurate documentation of the same conversation, in very different formats.

The solution

What we built

  • Compared transcription models on consultation audio before building on AssemblyAI.

  • Built plain-language visit summaries covering diagnoses, tests, medications and follow-up steps.

  • Added clinician formats, SOAP notes and referral letters, generated from the same transcript.

  • Managed and versioned prompts in PromptLayer, with a Django backend and LangChain orchestration.

Architecture

How the system works

How it fits together
  1. 01ConsultationRecorded visit audio
  2. 02TranscribeAssemblyAI, chosen by evaluation
  3. 03GenerateSummary, SOAP note, referral
  4. 04Review & sharePatients, families, clinicians
Engineering

How it was engineered

  • Chose the transcription model through a side-by-side evaluation on consultation audio, rather than by default.

  • Versioned every prompt in PromptLayer, so changes to medical outputs can be compared and rolled back.

Scope

What was delivered

01

Transcription model evaluation

02

Transcription pipeline

03

Visit summary generation

04

SOAP note and referral letter generation

05

Prompt management in PromptLayer

Technology

Stack

AssemblyAIOpenAILangChainPromptLayerDjango

Need something similar?

We can scope and build a system with comparable capabilities, shaped around your own workflows, data and integrations.