Healthcare Data Capture MVP
Budget: ₹12,500 – ₹37,500 INR
I’m putting together a demonstration MVP that can pull patient vitals, anaesthesia, and perfusion metrics straight off handwritten or printed charts while still allowing staff to key-in the same data manually when OCR falls short. The backend will live in FastAPI, the front-end in React / Next.js, and the OCR layer can rely on either Tesseract or AWS Textract—whichever you feel will get us from “grainy OR photo” to structured JSON fastest.
Here’s the core flow I need built:
• Upload or drag-and-drop screenshots or scans → run OCR → return a confidence score alongside parsed fields.
• Side-by-side manual entry screen so clinicians can confirm or overwrite values in real time. Both capture routes are equally important.
• Store everything in SQLite for the demo, but keep the schema Postgres-ready so we can flip the switch later.
• Multi-screen dashboard that refreshes live, showing the collected vitals, anaesthesia, and perfusion readings. I’ll need two alert channels: gentle reminders when fields stay blank too long and anomaly detection that flags out-of-range or conflicting numbers.
To show me you’re the right fit, send a quick note outlining:
– a link to any past dashboards or data-capture tools you’ve shipped,
– your preferred OCR approach for mixed-quality medical documents, and
– how you’d wire up real-time updates between FastAPI and React (WebSockets? polling?).
Acceptance criteria for delivery:
1. Working docker-compose bringing up FastAPI API, React dashboard, and SQLite DB.
2. Upload endpoint with OCR pipeline returning JSON plus confidence scores.
3. Manual entry UI merged into the same workflow.
4. Dashboard displaying live data with visible “data entry reminder” and “anomaly detected” banners.
5. Basic README covering setup, code structure, and where to swap SQLite for Postgres.
If that sounds like your wheelhouse, I’d love to review your approach and timeline.
Here’s the core flow I need built:
• Upload or drag-and-drop screenshots or scans → run OCR → return a confidence score alongside parsed fields.
• Side-by-side manual entry screen so clinicians can confirm or overwrite values in real time. Both capture routes are equally important.
• Store everything in SQLite for the demo, but keep the schema Postgres-ready so we can flip the switch later.
• Multi-screen dashboard that refreshes live, showing the collected vitals, anaesthesia, and perfusion readings. I’ll need two alert channels: gentle reminders when fields stay blank too long and anomaly detection that flags out-of-range or conflicting numbers.
To show me you’re the right fit, send a quick note outlining:
– a link to any past dashboards or data-capture tools you’ve shipped,
– your preferred OCR approach for mixed-quality medical documents, and
– how you’d wire up real-time updates between FastAPI and React (WebSockets? polling?).
Acceptance criteria for delivery:
1. Working docker-compose bringing up FastAPI API, React dashboard, and SQLite DB.
2. Upload endpoint with OCR pipeline returning JSON plus confidence scores.
3. Manual entry UI merged into the same workflow.
4. Dashboard displaying live data with visible “data entry reminder” and “anomaly detected” banners.
5. Basic README covering setup, code structure, and where to swap SQLite for Postgres.
If that sounds like your wheelhouse, I’d love to review your approach and timeline.