Nomi.ai — Your Intelligent Digital Health Twin

Story

A few months ago, I realized just how chaotic managing health the traditional way is. After getting a stubborn infection from a camping trip in Canada, and then later dislocating my knee and tearing a ligament, I began spending more and more time inefficiently researching symptoms and recovery methods.

I tracked everything in random places: symptoms in my Notes app, lab results buried in email threads, and even an ongoing ChatGPT chat that had practically turned into my medical diary. It worked — until my doctor asked, “So, walk me through what’s been happening lately?”

I had all the information, but no way to actually understand it as a whole. Despite all of the ways I stored data, I wasn't able to properly communicate or understand my current health state.


Problem

Our health shouldn’t depend on memory, luck, or a pile of disconnected notes. There has to be a better way to make sense of it all.

  • 70% of health decisions are made after symptoms appear
  • Over $300B is spent annually in the U.S. managing avoidable health issues
  • 60% of chronic conditions are preventable with early detection
  • There’s no singular system that makes personal health data intelligent and action-driven

Solution

That’s where Nomi.ai comes in. It's a digital twin that learns from everything your body’s been through, connects the dots you can’t, and gives you real, actionable insight. It’s not just storing your health, it’s actually understanding the reasoning behind it. It brings everything together, from doctor visits and scans, to your water intake and sleep, turning scattered information into one clear, connected view. Nomi helps you understand how your choices affect your well-being in real time and gives you intelligent feedback so that your health is back where it belongs, in your own hands!

Core Approach

Bringing everything together: We wanted all health data, like symptoms, sleep, activity, habits, and doctor notes, all in one place that actually makes sense together.

Making it smart: Instead of just storing information, Nomi uses AI to understand patterns, connect causes, and explain what might really be going on.

Keeping it natural: You can just talk to Nomi like a person. Ask questions, log how you feel, and get clear, relevant answers right away.

Our Presentation: https://www.canva.com/design/DAG07H5BoIw/w9uDFpEsJlIDpig-sBZetg/edit?utm_content=DAG07H5BoIw&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton


Features

AI Reasoning Layer: Uses medical context + daily metrics to answer questions like a clinician.
Conversational Interface: Talk to Nomi naturally — it listens, understands, and speaks back using ElevenLabs.
Symptom Insight Engine: Suggests causes, self-care, and follow-up guidance based on recent patterns.
Health Timeline: Tracks exercise, sleep, water, and nutrition for correlation with symptoms.
Real-time Voice Mode: Speak and listen hands-free with live transcript and responsive visuals.


Tech Stack

Backend

Node.js (ESM) – main runtime
Express 5.1 – lightweight API for chat + voice endpoints
dotenv 17.2.3 – loads .docenv + .env for secure API key access

Reasoning / AI

  • Anthropic SDK 0.65.0
  • claude-3-5-sonnet-latest
  • JSON schema enforced for every AI response ({ speak, next_q })

Speech & Voice

STT: Web Speech API (SpeechRecognition || webkitSpeechRecognition) — browser-side capture
TTS: ElevenLabs API (v1/text-to-speech)

  • Models: eleven_multilingual_v2, eleven_monolingual_v1
  • Voice ID stored in .docenv
    • Real-time voice synthesis + playback

Frontend / Chat UI

HTML, CSS, JS (served by Express)
EventSource (SSE): live response streaming
Visualizer: animated blob that reacts to speech amplitude
Transcript Panel: scrollable conversation history
Controls: Hold-to-talk, Pause/Interrupt, Audio playback toggle

Data Layer

Local JSON: patient_soumika.json

  • Includes demographics, conditions, medications, labs, symptoms, and the patients' daily health metrics In-Memory State:
  • state.prevDocQuestion
  • state.history (last 8 messages)

CLI Tools

  • doc_cli.mjs — terminal-based chat client with TTS MP3 output
  • tts_browser.js — Express speech API for browser playback

Next Steps

Integrations: Connect Apple Health, Fitbit, and Google Fit APIs
Doctor-Backed Predictive Analytics: Alerts providers about warnings for nutrition, stress, or sleep-related conditions to get immediate assistance Expanded Digital Twin: Full-body health representation with modular appendage tracking
Real User Trials: Gather anonymized user data for continuous improvement


Conclusion

Nomi.ai transforms everyday health data into actionable understanding.
It bridges the gap between information and insight and gives people the context they need to stay proactive, not reactive.

By combining reasoning, voice, and empathy, Nomi is a step toward a truly intelligent personal health companion!


Thank You

Nomi.ai Team, HackUTA 2025**

Built With

Share this project:

Updates