Products/Glyph
Glyph
Healthcare documents, read into structured data.
Glyph reads handwritten prescriptions, lab forms, reports and bills, then returns clean fields your team can check. Every value shows how sure the system is and where it came from on the page.
- Patient
- Sample PatientHidden
- Age · Sex
- 42 · FHidden
- Doctor
- Dr. Sample NameHidden
- Date
- 12 Oct 2026
3 fields hidden by your settings. Never sent to the model.
| Medicine | Frequency | Days | Confidence |
|---|---|---|---|
| Metformin500 mg · Tablet | 1-0-1 | 30 | 96% |
| Amlodipine5 mg · Tablet | 0-0-1 | 30 | 92% |
| Azithrol500 mg · TabletSuggestion: Azithromycin | 1-0-0 | 3 | Review · 58% |
Sample screen with synthetic data. Fields hidden in settings are masked and never sent to the model.
Paper still carries healthcare. Software needs data.
Prescriptions are handwritten, lab forms mix print and pen, and reports arrive as photos and PDFs in every layout. Typing them in is slow, and the errors that creep in are hard to spot afterwards.
Scribbled, abbreviated, rushed
Drug names, doses and durations written quickly, often crossed out or amended.
No two forms alike
Every clinic, lab and hospital lays out its documents differently, front and back.
Written the way clinics write
Dose patterns such as 1-0-1, OD and BD, duration shorthand and registration numbers follow Indian practice, and Glyph is tuned to read them.
Capture. Read. Check. Verify.
A web app that runs in any browser. Photograph or upload a document and review the result beside the original.
Camera or upload
Single pages or several together, including notes written on the back. Images are straightened and sized in the browser. HEIC and PDF are supported.
One structured read
The page is read into fields, each with a confidence score, a legibility rating and its exact position on the original.
Rules and a second look
Dose frequency, duration and route are normalised. Duplicate therapy, Schedule H1 drugs and missing doses are flagged. Uncertain lines are re-read with deeper reasoning.
A person confirms
Click a line to see its box on the page, edit any field in place, accept a suggestion and mark each line verified. Then export or print.
One reader for the documents healthcare runs on.
Prescriptions and lab requests have dedicated layouts and rules. The other types are configured and are being validated with partners. Adding a type is a configuration change, not a rebuild.
Clinical
- Prescription
- Discharge summary
- Referral letter
- Medical certificate
- Vaccination record
Diagnostics
- Lab test request
- Lab report
- Radiology report
Administrative
- Bill or invoice
- Insurance claim or card
- Consent form
- Other healthcare document
A transcriber, not a clinician.
Glyph turns what is on the page into data and says how sure it is. It never decides what the page should have said, so a person stays in charge of every value that matters.
- Confidence on every lineLow-confidence and scribbled lines are flagged, never hidden.
- The source one click awayEach field links to its box on the original, with a zoomed view of uncertain handwriting.
- Suggestions, not rewritesLikely drug-name corrections are offered for review and never applied silently.
- Clear limitsGlyph does not check drug interactions or dose safety, and says so in the product.
Show only what you need to read.
Each deployment is configured for the documents and fields an organisation handles, with its own name and settings.
- Choose document typesTurn types on or off. A type that is off is refused.
- Hide personal detailsPatient name, age, doctor or any other field can be hidden. A hidden field is never sent to the model and is removed from results even if returned.
- Edit in placeFix any value on the main screen. Add or remove rows in tables such as lab results and bill items.
- Take the data with youExport CSV or JSON, or print a clean copy for the record.
Production-grade, from day one.
Glyph runs as a dedicated, access-controlled service for each organisation, and verified results flow into your production database, ready for the systems that use them.
Your production database
Verified results are written to your organisation's database, so downstream systems work from clean, structured records.
Access controlled
Every request needs a verified identity through your identity provider. If access control isn't configured, Glyph refuses to run.
Minimal by default
Aadhaar numbers are never extracted, and fields you hide are never sent to the reading model.
Dedicated deployment
Each organisation runs on its own deployment, with its own document types, fields and settings.
What teams ask us.
Bring a few of your own documents and we'll walk through the result together.
How accurate is it on handwriting?
It depends on the writing. Rather than hide uncertainty, Glyph scores every line, flags low-confidence ones and lets a person verify each value. We recommend keeping that verification step for anything clinical.
Which documents can it read?
Prescriptions and lab test requests first, with lab reports, radiology reports, discharge summaries, referral letters, certificates, vaccination records, bills, insurance claims and consent forms also configured. Anything else can be read as a general healthcare document.
Can we add our own document type?
Yes. A document type is a set of fields and, where needed, a table of rows. Adding one is a configuration entry, not a new build.
Does it check drug interactions or doses?
No. Glyph reads and structures what is written. It flags duplicate therapy, Schedule H1 drugs and missing doses for review, but it makes no clinical decisions.
Where does it run?
Glyph runs as a secure web app behind your identity provider, with results saved to your production database. Staff need only a browser or a phone camera.
Turn paper into data your team can trust.
Tell us which documents you handle. We'll show you what Glyph reads, how it flags doubt, and how review works.