Engineer reviewing an OCR engine and document pipeline on multiple screens

Lab

Ideas, prototypes and technical experiments I'm exploring.

LAB / 001

BUILDING

Medical Document Intelligence

Document → OCR → AI extraction → timeline → human review.

Try the experiment
ASP.NET CoreReactOCRLLMs

LAB / 002

EXPERIMENT

Receipt Intelligence

Turning ordinary receipts into structured household purchase data.

OCRVisionData extraction

LAB / 003

RESEARCH

OCR Reliability Lab

Measuring OCR accuracy, AI correction and structured extraction reliability.

OCRAIResearch
DOCUMENT INTELLIGENCE LAB PREPARED SAMPLES

Watch a document become structured data.

Pick a synthetic sample document and run it through the pipeline: OCR reads the page, AI extraction turns it into structured fields, and anything below the confidence bar gets flagged for human review.

Green Valley Clinic
Dr. S. Karim, MBBS · Reg 48211
Date: 12 May 2026
Patient: A. Rahman, 54M

1. Amlodipine 5 mg
Sig: 1 tablet once daily, morning — 30 days

2. Metformin 500 mg
Sig: 1 tablet twice daily with meals — 30 days

Review blood pressure in 4 weeks.

Signature: S. Karim  ·  Next visit: 09 Jun 2026

Prepared sample documents with synthetic patient data — nothing here is real, and this demo doesn't accept uploads. The results replay output from the same OCR → extraction approach the project explores, with low-confidence fields routed to human review instead of being trusted automatically.

CONTACT

Curious about one of these?

These experiments are open to collaboration, feedback or a serious build.

Start a conversation me@sabitbrinta.com

Usually replies within a day