DocAssist-AI

Clinic visits often begin with repeated questions, language barriers, and rushed intake. Doctors lose time collecting basic details, and patients with urgent symptoms can wait in the same queue as everyone else. DocAssist AI makes this first step safer, clearer, and faster for both patients and doctors. A patient picks a language, gives consent, and describes their symptoms to an AI that asks one short question at a time. The AI flags possible emergencies, suggests the right type of doctor, and books a queue token. The doctor then sees a structured summary with a confidence badge, so they know how far to trust it, and reviews every AI suggestion, approving or rejecting each one and giving a reason when rejecting. A medicine check compares the medicines the patient mentions with the doctor's list and marks anything only the AI suggested. After treatment, the patient sees the doctor's approved advice and can rate the visit. Safety is built into every step. The AI never gives a diagnosis or names medicines in its chat replies, and its suggestions appear only after the conversation ends and stay pending until a doctor decides. If a patient describes a possible emergency, the intake stops and the app tells them to get help now. Patients give consent before any data is entered, and our demo uses fake data only. Because a doctor approves or rejects everything, nothing reaches the patient without human review, and an accuracy panel tracks how often doctors agree with the AI. On our test cases, the emergency check recorded 0 missed emergencies and 0 false alarms. DocAssist AI is built with React, Vite, and Node.js. The AI runs on Gemma-4-31B-IT served through AMD Token Factory.