Building a Scalable AI-Powered Transcription System for Healthcare
Project Year
2022
Industry
Software Development
Overview
A UAE-based software house aimed to develop an AI-powered healthcare solution for multiple hospitals. The product’s core functionality was to transcribe patient-doctor sessions, generate structured summaries, and store the data securely for future reference. The backend was developed using NestJS, while the frontend was built with Next.js.
The system leveraged Azure Transcription to process audio recordings and incorporated a form for doctors to input physical examination details. The combined data was then processed using ChatGPT to generate summaries and insights. The project evolved to support multi-tenancy, feature flagging, real-time monitoring, and HIPAA compliance.
Client Challenge
Real-Time Transcription Complexity The initial approach of processing audio chunks in real time led to memory leaks and high latency.
Multi-Tenancy Implementation The system needed to support multiple hospitals while maintaining data segregation.
User Management Migration The client introduced an external system for user management, requiring a seamless migration without data loss
Compliance with HIPAA Regulations Patient data security had to meet strict healthcare standards.
Performance & Monitoring Ensuring smooth system performance and quickly identifying issues was critical.
Solutions Approach
Initially, audio chunks were sent from the frontend to the backend, which then forwarded them to the transcription service over WebSockets. However, this created performance bottlenecks.
Removed Backend from the Real-Time Pipeline: Audio chunks were sent directly to the transcription service.
Efficient Data Flow:The transcription service processed audio in real time and stored the full audio in AWS S3.
Latency Reduction:Once the session ended, the complete transcription and audio URL were sent to the backend, improving efficiency.
Agile Development & Collaboration
01
Daily Scrum Meetings
Developers shared updates and discussed blockers.
02
Weekly Backlog Grooming
Tasks were refined based on evolving requirements.
03
Sprint Planning & Reviews
Fortnightly sprint planning with effort estimation.
04
Strict Code Review Process:
Feature branches were named after JIRA tickets.
Key Features
Enhanced Performance: Optimized transcription flow reduced memory improved system responsiveness.
Improved Trust Hospitals could manage their users independently while maintaining data security.
Scalability & Flexibility:Feature flagging enabled hospitals to customize services easily.
Improved Compliance: Secure data storage and HIPAA compliance ensured regulatory approval.
Better System Observability: Datadog and SonarQube helped maintain high reliability and performance.
Accelerated Development: AI-powered tools reduced coding time and improved overall software quality.
Conclusion
This case study demonstrates how AI-driven automation, secure data handling, and scalable architecture can transform healthcare transcription and record-keeping, improving efficiency and compliance in medical institutions.