Nirog

Inspiration AYUSH doctors -practitioners of Ayurveda, Yoga, Unani, Siddha, and Homeopathy -handle large patient loads, especially in rural and semi-urban clinics, yet still rely on manual, paper-based case-taking. Every minute spent writing down a complaint by hand is a minute not spent treating someone. Patients also routinely forget to take prescribed medicine on time, with no system to remind them. We spoke directly with a practicing doctor about this idea, who confirmed the problem - and pushed us further, asking for something that could genuinely help with diagnosis, not just intake.

What it does (planned)

Nirog is designed as an end-to-end digital case-taking system for AYUSH clinics:

A patient will enter their details and describe their complaint — by typing or speaking directly into the app. An AI model will read the complaint and generate a structured summary, extracted symptoms, and smart follow-up questions, plus a short list of possible conditions for the doctor to consider - never a diagnosis, always framed as discussion points for a qualified professional. The patient will be able to answer those follow-ups directly in the app, refining the case before a doctor sees it. A doctor will open their dashboard, review the AI-generated summary and possible conditions, examine the patient, and record vitals and a diagnosis. The doctor will prescribe medicine with dosage and timing, and schedule a reminder. At the right time, the patient will receive a WhatsApp message reminding them to take their medicine.

How we plan to build it

Backend: Python + FastAPI, with REST endpoints for patients, doctors, case sheets, prescriptions, and reminders. Database: Supabase (PostgreSQL) for structured data. AI: Google Gemini for extracting structured case data and suggesting possible conditions from free-text complaints. Voice input: Browser-native speech recognition for hands-free complaint entry. Frontend: React + Vite + Tailwind CSS — a patient-facing flow and a separate doctor dashboard. Reminders: A background scheduler triggering real WhatsApp messages via the Twilio API. Challenges we anticipate

Keeping AI prompts medically cautious — surfacing possible conditions as discussion points without ever implying a confident diagnosis. Matching our data model cleanly to a real clinical workflow (patients, case sheets, prescriptions, reminders) without over-engineering before we've built anything. Reliable WhatsApp delivery through Twilio, including sandbox opt-in constraints for test numbers. Designing for genuinely low-resource environments — older hardware, patchy internet — which shapes choices like favoring lightweight browser-based voice input over heavier server-side transcription.

What we've accomplished so far

A complete architecture and technical plan, covering every step from patient intake to doctor review to medicine reminders. Direct validation from a practicing doctor, including specific feedback that led us to add AI-assisted "possible conditions" to the plan. A clear, buildable tech stack with no unresolved dependencies.

What we learned

Talking to an actual doctor early reframed the problem for us — intake speed matters, but clinical decision support is where the real value is. A clear blueprint, grounded in real user feedback, is worth more at this stage than a half-built feature list.

What's next for Nirog

Build the core flow: patient intake → AI extraction → doctor review. Implement the examination and diagnosis step for doctors. Wire up real WhatsApp reminders end-to-end. Get the finished build back in front of the doctor we consulted for a second round of feedback.