How AI Reads a Prescription — and Turns It Into a Medicine Basket

Talya Medical's OCR prescription reader runs on-device and offline, backed by a drug-data engine with a strict no-fabrication rule. Here is how it works and why safety comes first.

Turning a photo of a handwritten prescription into an accurate medicine list is a hard — and sensitive — task. A single error can mean the wrong drug. That is why we designed Talya Medical’s prescription reader around two principles: accuracy and no fabrication.

Stage one: on-device reading (OCR)

The pharmacist takes a photo of the prescription, and the optical-character-recognition engine extracts the text locally on the device — without uploading any image to the internet. This protects patient privacy and keeps it working even without a connection.

Stage two: understanding with a “no-fabrication” rule

The extracted text is passed to the drug-data engine powered by advanced AI models, under a strict rule: the engine invents nothing. If a drug is not confidently known, confirmation is requested instead of a guess. In medicine, “I don’t know” is safer than a fabricated answer.

Stage three: clinical safety

Before dispensing is completed, the clinical-safety assistant checks:

Drug interactions between basket items.
Recorded patient allergies.
Known dosages and warnings.

Why this approach?

Many “AI solutions” promise magic and then produce unreliable information. We prefer a conservative, auditable intelligence: it assists and speeds up the pharmacist, but it neither replaces their judgment nor invents data.

Good medical AI knows its limits — and asks for confirmation when in doubt.

Try the prescription reader inside PharmaSys and request a demo.