HEALTGURU

Inspiration: Take Ramesh, 58, who has diabetes and high blood pressure. He sees three doctors, and each one writes a prescription without knowing what the others prescribed. India has no nationalized Electronic Health Record, so every specialist keeps an isolated record. The published median prescribing-error rate is 34.11%, and adverse drug events lead to 56% longer hospital stays. No one sees the full picture except the patient's body. We wanted to build the check that no single doctor can make alone.

What it does:

HealthGuru reads any prescription or lab report, including handwritten ones. It checks the medicines against everything else the patient is taking and every lab value on record. It then gives a Safety Score and alerts both the patient and the doctor. When it finds a risky combination, it can also suggest a cheaper generic. In our example, a ₹4,500 branded regimen comes down to about ₹850 at a Jan Aushadhi kiosk.

How we built it:

Scan and digitize: computer vision reads handwritten prescriptions and lab reports. Parse and verify: the text becomes structured fields. Anything below 95% confidence goes to a manual check. Map to chemical salt: a localized Indian brand-to-salt index turns brand names into their active salts. Dual contradiction check: a Salt × Salt matrix catches drug–drug interactions. A Salt × Lab Biomarker matrix catches drug–lab conflicts. Safety Score and alert: the two checks merge into one score, shown to the patient and the doctor. Challenges we ran into Reading messy handwriting reliably. Matching the many Indian brand names to the right chemical salts. Getting lab data in a usable form. Deciding when the system should trust itself and when it should hand off to a human. This is why we added the 95% confidence rule. Handling patient privacy and consent for sensitive health records. Accomplishments that we're proud of A dual-matrix approach that checks drugs against labs as well as against other drugs. Single-platform pharmacy systems don't make this distinction. A design that works across doctors, so it isn't limited to one pharmacy's sales data. A working end-to-end flow from scanned prescription to alert. (Add your real results here, such as OCR accuracy or the number of interactions covered.) An early market estimate of about $45M obtainable over 3 years. This is preliminary sizing, not a final figure.

What we learned:

Most prescribing errors come from missing information rather than from careless doctors. Healthcare software needs a human in the loop wherever confidence is low. Local context matters. A safety tool for India has to understand Indian brand names and Indian generics. A clear, simple alert is worth more than a long list of warnings. What's next for HealthGuru Integrate across providers, so records connect wherever the patient goes. Expand branded-to-generic mapping. Partner with clinics, diagnostic labs and insurers. Capture the roughly $45M market.