Ninx

What it does

The app takes input of an image of a medical document containing a disease and/or treatment, it extracts this information using Gemini's vision capabilities and then generates detailed information with Gemini in non medical terms to give, for the disease: cause, signs and symptoms, treatment and prevention and for the treatment: generic name, indications, mechanism of action, side effects and contraindications. This information is then presented back to the user.

We are focusing on images to preserve the information as obtained from the attending health personnel and also increase accessibility to people who can't read but can use a smartphone. Our primary target are communities still using paper medical records, specifically Uganda since we know there is need there.

Built to be an add-on to visiting a doctor, in case the patient wants to find more information about the disease and/or treatment and not intended in anyway to replace the clinician.

We acknowledge the limitations of the application of LLMs in healthcare (especially hallucinations in both text and OCR work), and clearly state these on all application webpages except the landing page.

Although Gemini 1.5 Pro is more accurate in OCR work, since we're serving areas that can't pay for the service, it's more sustainable to use 1.5 Flash although it has shown reduced accuracy, as expected. We are optimistic that if Gemini continues to improve as has been the trend, Ninx will also improve automatically.