RespondrAI

Client Snapshot

• Industry: Healthcare (telemedicine and emergency services, US)
• Client: Tele911
• Headquarters: Houston, Texas, US
• Timeline: August 2025 to launch in December 2025
• Team: 1 AI Architect · 1 Backend Engineer · 1 Mobile Engineer · 1 Project Manager
• Services: AI Solutions · Mobile App Development · Platform Development · Cloud/DevOps
• Stack: Claude (Anthropic) · AWS Bedrock · AWS Transcribe · React Native + Expo · NestJS · PostgreSQL · Socket.io · AWS GovCloud (US) · Docker · Sentry
• Traction: Live on the App Store and Google Play; about 750 SOAP notes drafted in the 90 days to October 2026

The Problem

After every visit, clinicians wrote a structured SOAP note by hand before they could move on to the next patient.

The problem wasn't recording the visit. It was turning a conversation into a clinical record a doctor can sign.

• Documentation load: every visit ends with a SOAP note to write.
• Who said what: a usable note depends on telling the doctor's words from the patient's.
• Structured patient data: name, date of birth, age, sex and reason for visit have to land in the right fields, every time.
• Healthcare data rules: recordings and notes need a HIPAA-compliant environment.
• Existing sign-in: clinicians already work in VSee, so the app had to fit into it.

‍

The Solution

We built RespondrAI as an event-driven pipeline: the app records, the cloud transcribes and drafts, and the note comes back to the clinician's phone in real time.

AI Pipeline

• Speaker-separated transcription: AWS Transcribe converts the audio and labels doctor and patient.
• Claude on AWS Bedrock: Claude 3 Haiku drafts the SOAP note and extracts the patient data, both in parallel.
• Event-driven flow: S3, Transcribe, SNS and SQS feed a NestJS consumer, which notifies the app over WebSocket when the note is ready.
• Prompts as data: prompt templates live in the database, so the team tunes them without a release.
• Audit trail: every SOAP note is versioned.

Mobile and Platform

• One codebase, two stores: React Native and Expo apps, signed and published on the App Store and Google Play.
• Secure API: NestJS on Node.js 22, OAuth sign-in through VSee, and our own JWTs signed with RS256.
• Real time: Socket.io connections authenticated with JWT.
• Shareable notes: PDF export from a dedicated Markdown-to-PDF service.

Compliance and Cloud

• AWS GovCloud (US): S3, Transcribe, SQS, SNS, SES and RDS for clinical data.
• Locked-down storage: pre-signed URLs for audio, transcripts and PDFs; TLS on every database connection.
• Monitoring: Sentry on the API.

‍

AI Core

SOAP note drafting

Claude drafts a structured SOAP note from every recorded visit, ready for the clinician to review.

Patient data extraction

Name, date of birth, age, sex and reason for visit pulled from the conversation into structured fields.

Doctor and patient separation

Transcription that labels each speaker, so the note reflects who said what.

______
‍

Infrastructure

GovCloud data layer

Clinical data stored in AWS GovCloud (US), on a HIPAA-compliant architecture.

Event-driven pipeline

Upload, transcription and drafting run as queued events, with the app notified the moment the note is ready.

______
‍

Product

Clinician app

iOS and Android app that records the visit and shows the note in real time.

‍

Results

• Notes drafted automatically: every recorded visit becomes a draft SOAP note.
• Patient details captured: five core fields extracted without typing.
• Live in both stores: published on the App Store and Google Play.
• HIPAA-compliant: clinical data in AWS GovCloud (US), with versioned notes as an audit trail.
• Tunable without releases: prompts managed in the database.
• In daily use: about 750 SOAP notes drafted in the 90 days to October 2026, roughly 8 a day.

‍

KPIs

2 app stores — iOS and Android from one React Native codebase, published and signed

5 patient fields — Name, date of birth, age, sex and reason for visit, extracted by Claude

~750 notes — SOAP notes drafted in the 90 days to October 2026, about 8 a day

‍

Tech Stack

• AI: Claude 3 Haiku on AWS Bedrock (Converse API), AWS Transcribe with speaker diarization
• Mobile: React Native, Expo (Expo Router), expo-av
• Backend: Node.js 22, NestJS 11, TypeScript 5.7, PostgreSQL with TypeORM, Socket.io, PDFKit
• Cloud: AWS GovCloud (US): S3, Transcribe, SQS, SNS, SES, RDS; Amazon Bedrock
• DevOps: Docker, AWS SDK v3, Sentry

‍

Team & Delivery Model

A four-person senior squad:

AI Architect / Tech Lead

Transcription and drafting pipeline, prompt design, GovCloud architecture.

Backend Engineer

NestJS API, event consumers, VSee integration.

Mobile Engineer

React Native app, audio capture, store releases.

Project Manager

Scope, milestones and client communication.

Operating Model

Agile, milestone-driven delivery with iterative releases.

‍

Claude Partner Network · Anthropic Services Partner Directory · Claude Certified Architect. Built on Claude by a Claude Partner Network team.