AADITHYA VIMAL

Software · 2026 · Live demo · Open source · 2026

Axiosync

A proactive AI health companion featuring an ultra-realistic 3D human anatomy engine with muscular exhaustion heatmaps and 100% on-device WebLLM inference.

My role
Designed and built the Three.js 3D anatomical model, WebLLM client inference engine, and Firebase synchronization layer.
Status
Live demo · Open source · 2026
Category
Software · Proactive AI health companion
Live demoSource on GitHub

Live demo · interactive

🔒axiosync.pages.dev
CONNECTING TO AXIOSYNC...

Embedded for convenience — if it doesn’t load here, open the full site directly.

01

Understand

Why it exists. Passive health loggers record numbers without illustrating physical consequences, while cloud AI coaches create severe privacy risks with biometric records.

What I built. Proactive wellness platform combining an interactive 3D human anatomy heatmap with a 500+ bodyweight exercise workout engine and a 100% on-device AI health coach powered by SmolLM2-1.7B running in browser Web Workers.

The problem it solves. Conventional health trackers act as passive logbooks that fail to visualize how daily choices impact physical systems, while cloud-based AI health coaches present significant privacy concerns by sending biometric records to remote model servers.

What it does

  • Interactive 3D anatomical model mapping muscular exhaustion, lung stress, and liver burden in real time
  • Smart Workout Engine with 500+ bodyweight exercises and rolling 7-day training volume tracking
  • SmolLM2-1.7B running locally in browser Web Workers for private AI health coaching
  • Multifaceted tracking covering nutrition, circadian sleep windows, and 12+ sport activities
02

Technical depth

Architecture. Three.js WebGL anatomical model renders dynamic hotspots for muscular exhaustion, lung stress (smoking logs), and liver burden (alcohol logs). Calculates rolling 7-day training volume, circadian sleep schedules, and logs 12+ sports activities, synchronizing profile data to Firebase Firestore.

Security model. Client-side AI inference safeguards confidential health information: SmolLM2 runs locally via Web Workers with zero biometric telemetry sent to external AI servers. Google OAuth authentication with user-controlled data erasure.

Tradeoff. Running 3D WebGL rendering alongside client-side LLM inference requires modern browser hardware, but guarantees biometric privacy.

  • React
  • TypeScript
  • Firebase
  • Convex
  • Tailwind CSS
How it works, step by step

Three.js WebGL anatomical model renders dynamic hotspots for muscular exhaustion, lung stress (smoking logs), and liver burden (alcohol logs). Calculates rolling 7-day training volume, circadian sleep schedules, and logs 12+ sports activities, synchronizing profile data to Firebase Firestore.

Security notes

Client-side AI inference safeguards confidential health information: SmolLM2 runs locally via Web Workers with zero biometric telemetry sent to external AI servers. Google OAuth authentication with user-controlled data erasure.

Software