AI Automation · Ongoing — Oxibit Technologies
AI-Powered Clinical Visit Documentation System
Sigma Rev is an AI-powered clinical productivity suite that turns patient visit conversations into structured documentation — clinical notes, after-visit summaries, diagnoses, procedures, and real-time copilot suggestions — so clinicians spend less time on paperwork and more time with patients.
Role: Full-Stack Engineer — Frontend & Integration
Next.js 15React 19TypeScriptDjango 5Django REST FrameworkOpenAIMySQL
Real-time browser audio capture via WebSocket
Recording
5 AI-generated document types per visit
Documentation
Pending → Reviewed → Edited → Submitted
Workflow
The Problem
Clinical documentation is one of the biggest bottlenecks in healthcare — clinicians often spend hours after each visit writing notes, coding diagnoses, and preparing patient summaries, and existing EHR workflows are slow, fragmented, and pull attention away from care. Sigma Rev was built to capture visit audio in real time, transcribe and diarize the doctor-patient conversation automatically, generate structured clinical outputs with AI, and give clinicians a clear review-and-submit workflow before anything goes into the record.
- Capture visit audio in real time during appointments.
- Transcribe and diarize doctor–patient conversations automatically.
- Generate structured clinical outputs with AI.
- Give clinicians a clear review-and-submit workflow before documentation goes into the record.
The Approach
- 01Built visit worklists — completed and upcoming visit views with search, filters, pagination, and one-click pre-visit summaries.
- 02Built the live recording interface: browser-based audio capture via the Web Audio API and a custom AudioWorklet (PCMProcessor) streaming 24 kHz mono PCM to the backend over WebSocket, with live diarized transcription and AI copilot suggestions.
- 03Built the post-visit review UI with tabbed navigation across five AI-generated document types (clinical notes, after-visit summary, diagnoses, procedures, copilot).
- 04Implemented the review workflow: scroll-based 'mark as reviewed' behavior and status tracking through Pending → Reviewed → Edited → Submitted.
- 05Built per-clinician note customization — section formatting preferences and which of 8 standard clinical sections appear by default.
- 06Implemented the authentication flow, including login with optional OTP email verification.
Key Features
Visit Worklists
- Completed Visits: searchable, filterable table (patient name, DOB, MR number, date range) with review-progress status indicators.
- Upcoming Visits: scheduled appointments with one-click Pre-Visit Summary (conditions, medications, abnormal findings, prior notes) and Start Recording.
Real-Time Visit Recording
- Microphone capture via the Web Audio API and a custom AudioWorklet streaming 24 kHz mono PCM to the backend over WebSocket.
- Live transcription with doctor/patient speaker diarization, AI copilot follow-up suggestions, and pause/resume/stop controls with a live timer.
AI-Generated Clinical Documentation
- Clinical Notes (HPI, PMH, Medications, ROS, Physical Exam, Assessment & Plan), After-Visit Summary, Diagnoses with ICD codes, Procedures with codes and rationale, and Copilot suggestions — each its own review tab.
Clinician Review Workflow
- Each section moves through Pending → Reviewed → Edited → Submitted with status badges and timestamps.
- Scroll-to-review marking, copy-to-clipboard for EHR paste, and a sidebar transcript viewer with audio playback for cross-checking AI output.
Personalized Note Settings
- Per-clinician configuration for HPI style (concise vs. comprehensive), Assessment & Plan format (bulleted vs. paragraph), and which of 8 standard sections appear by default.
Technical Highlights
Frontend Architecture
- Next.js 15 App Router with route groups for auth and protected pages.
- React Context (UserContext, VisitContext) with localStorage persistence for session and visit state.
- Axios with CSRF token handling for Django session auth, plus a reusable table pagination, modal system, and custom SVG icon set.
Backend Integration
- Django REST Framework endpoints for CRUD, filtering, and status updates, with Django Channels handling the WebSocket audio stream.
- OpenAI for transcription, diarization, and content generation, with flexible audio storage (local filesystem or Dropbox) and optional OTP email verification for login.
UX Decisions
- Three-column visit layout: patient info, document content, and live transcript side by side.
- Color-coded status pills on the worklist, a pre-visit summary offcanvas, and mobile-responsive table columns with abbreviated labels on smaller screens.
My Contribution
- Built visit management: completed and upcoming visit worklists with search, filters, and pagination.
- Built the live recording interface — browser-based audio capture with real-time transcription over WebSockets.
- Built the post-visit review UI with tabbed navigation across five AI-generated document types.
- Implemented the review workflow — scroll-based 'mark as reviewed' behavior and status tracking (Pending → Reviewed → Submitted).
- Built note customization — per-clinician settings for note sections and formatting preferences.
- Implemented the authentication flow, including login with optional OTP email verification.
Tech Stack Summary
FrontendNext.js 15, React 19, TypeScript, Tailwind CSS 4, HeroUI, Axios, Framer Motion, Sonner, Moment.js
BackendDjango 5, Django REST Framework, OpenAI API, MySQL, Dropbox, Django Channels (WebSocket)
InfrastructureSession-based auth with CSRF, optional OTP email verification, environment-driven config (test vs. production AI modes)