InstantHPI — Free AI Medical Consultations

Free AI doctor on Telegram. $0.003/consult. 20+ languages. Try: t.me/instanthpibot

Comment

Inspiration

3.5 billion people have never seen a doctor. Not because they don't want to — because there isn't one. In Nigeria, 5,500 people share one doctor. In Ethiopia, it's 10,000 to 1. Even where doctors exist, they spend 15-30 minutes per patient just on documentation.

As a Canadian family physician, I asked: what if AI could handle the documentation so doctors could focus on patients?

What it does

InstantHPI is a Telegram bot (@instanthpibot) that conducts a complete medical consultation in 2 minutes. A patient describes their symptoms in any of 20+ languages. The AI asks structured follow-up questions — just like a physician would — then generates:

  • Complete History of Present Illness (HPI)
  • SOAP clinical note
  • Differential diagnosis with clinical reasoning
  • Prescription template
  • Referral letter
  • Lab and imaging recommendations
  • All as downloadable PDFs

Cost per consultation: $0.003. Patient pays: $0.00. Always.

How we built it

  • AI Engine: Claude (Anthropic) via AWS Bedrock — HIPAA-compliant infrastructure
  • Platform: Telegram Bot API — 900M+ users, works on basic smartphones, no app install needed
  • Backend: Node.js on AWS EC2
  • Documents: PDF generation with structured clinical templates
  • Languages: Auto-detected, 20+ supported via AI translation to clinical English

Challenges we ran into

Getting AI to conduct a structured medical interview that mirrors how a real physician takes a history — not just answering questions but actively probing for critical details. Ensuring safety: every output includes urgency flags, disclaimers, and "see your doctor" recommendations. Supporting 20+ languages while maintaining clinical accuracy.

What we learned

AI doesn't need to replace doctors — it needs to replace the notepad. The documentation burden is the real bottleneck. At $0.003 per consultation, the cost barrier to healthcare access is effectively zero.

What's next

  • Piloting in Canadian emergency departments (patients use bot in waiting room)
  • Applied to MIT Solve ($250K), CHISEL Singapore (S$500K), Gates Grand Challenges
  • Contacted WHO, UNICEF, MSF, Gates Foundation, and health ministries in 6 countries
  • 78 emails delivered to organizations across 50+ countries

Try it now: https://t.me/instanthpibot Case study: https://instanthpi.com/case-study.html Our journey: https://instanthpi.com/story.html

Built With

  • ai/ml
  • aws-bedrock
  • claude-ai
  • healthcare
  • node.js
  • telegram-bot-api
Share this project:

Updates