An on-premise AI platform that reads unstructured clinical documents — scanned notes, handwritten prescriptions, lab reports — and turns them into structured, source-linked, research-ready data. Nothing ever leaves the hospital network.
So instead of drawing the brachial plexus in our notebooks, we started sketching system architecture.
We walked in expecting stethoscopes and scrubs. We got mountains of paper instead — handwritten records, illegible prescriptions, lab reports stuffed into folders that no one would ever open again.
Between Anatomy lectures and Biochemistry vivas, we kept asking each other the same question — and we couldn't let it go.
Three medical students. One conviction: clinical records should never be a barrier to better care.
Before a study can even begin enrolling, someone has to read it all by hand.
Manual chart abstraction from scanned notes, handwritten prescriptions and free-text discharge summaries — for every single record.
Multiplied across a 500-patient study, that's hundreds of hours of pure data entry before the first patient is enrolled.
Different abstractors read the same chart differently — quietly threatening the integrity of multi-site research data.
A fully local pipeline that converts unstructured clinical documents into structured, research-ready data — every step on the hospital's own servers.
Not a roadmap. Five live modules you can open right now.
Real-time pipeline overview with document upload and processing status.
Automatically aggregates lab trends and medications across disconnected reports.
Generates clinical synopses and surfaces ranked safety signals.
OCR quality gate with auto-flagging and an audit log tracking every event, model & confidence score.
Powered by local Llama 3.1 to query across all uploaded documents securely.
Click through all five modules with live mock clinical data.
| Cloud Clinical AI | DOCUMENTITIS | |
|---|---|---|
| Data location | Sent to cloud | Stays in hospital |
| Infrastructure | Data-center GPUs | Consumer hardware |
| Traceability | Black-box outputs | Source-linked, audit-grade |
| Regulatory fit | Difficult in India / EU | Built natively for data sovereignty |
| Cost to hospital | High recurring fees | One-time deployment |
Automates chart abstraction for studies on sepsis, readmission rates and oncology outcomes — turning months of manual review into hours.
Standardises extraction across hospitals, eliminating the site-to-site inter-rater variability that compromises trial integrity.
Continuously evaluates records to flag abnormal lab values and safety signals — supporting real-time pharmacovigilance.
We're three medical students with a working v1.0.0. Here's where the right institution changes everything.
A clinical research institution willing to host a pilot deployment, strictly on de-identified historical data.
Permission to validate our extraction accuracy against ground-truth data, under proper ethical clearance.
Clinical research expertise to help us refine which use cases to prioritise first.
Clinical records should never be a barrier to better care.