AI models & experiences for wellness companionship
AI companions for change that takes root.
Root Wellness Labs designs and trains open-source wellness companion models — language, agent and vision — that run privately on your own computer or a lightweight home server. Your health data stays yours while you eat better, move more and stay accountable with a small community.
The problem
The most personal data deserves the most private AI.
Real wellness coaching needs real context: your goals, progress photos of your body, bloodwork results, medical history. That is exactly the data people are most reluctant to hand to a cloud service — and least willing to see turned into advertising.
So people either share too little to get useful help, or share it and lose control of it. We think there's a better option: purpose-trained, open-source models that live on hardware you own, remember your goals, and never need to send your data anywhere.
Products
Our models, and the experience they power
Three open-source model families, each trained for one job and sized to run at home, working together inside one companion experience.
Root Companion
Language models fine-tuned for wellness coaching conversations — specific, encouraging and honest about uncertainty.
Open weights · in developmentRoot Vision
Vision-language models that understand meal photos and exercise form, replacing tedious manual logging.
Open weights · in developmentRoot Agents
Accountability agents that plan your week, schedule check-ins, track commitments over months and coordinate your circle.
Open weights · in developmentThe Companion
The consumer experience: daily check-ins by text, voice or photo, personal insights and small peer circles — powered by all three models, running on your own computer or home server.
Pilot Q1 2027 · launch Q2 2027What the companion covers
Six domains our models are trained and evaluated on — where small, consistent changes compound.
Technology
Open models, trained on GPUs, built to run at home
We combine frontier AI infrastructure experience with the fast-growing open-source model ecosystem. We train on NVIDIA RTX GPUs, then optimize every release to run on the consumer GPUs and lightweight servers people already own.
-
01 · Data
Curated wellness corpora
Coaching dialogues, nutrition and exercise literature, and labeled meal and movement imagery, gathered with consent.
-
02 · Train
Fine-tuning on our GPUs
Open foundation models adapted for coaching, vision and planning on NVIDIA RTX hardware, scaling to RTX PRO-class GPUs.
-
03 · Evaluate
Safety-first benchmarks
Every release is tested for accuracy, tone and safe escalation — the companion knows when to point people to a professional.
-
04 · Deploy
Runs on hardware you own
Compressed and optimized open weights for home computers and lightweight servers, so photos, labs and conversations never have to leave the house.
How we build
Principles behind every release
Community over content
People change alongside people. Our models exist to keep small groups connected and consistent.
Evidence before hype
Trends are weighed against the research, and the companion says plainly when the science isn't settled.
Private by design
Your health data stays on your hardware. No ads, no data sales — open models you can run, inspect and trust.
Support, not substitute
The companion complements doctors and coaches — it never diagnoses, prescribes or replaces professional care.
Company
About Root Wellness Labs
Root Wellness Labs was founded by Samuel Hsiung to answer a simple question: how do you get truly personal wellness guidance without giving up your most personal data?
Our answer is to design and train open-source models that people run locally, on a home computer or a lightweight server. We bring years of experience at the frontier of AI technology and infrastructure to the open-source ecosystem, and turn it into wellness companions people can actually trust.
- Headquarters
- Greater Seattle, WA
- Incorporated
- December 2025
- Focus
- Open Sourced On Device Wellness AI
Team
Samuel Hsiung · Founder & CEO
Ex-Google, ex-Meta. Data scientist and technical program manager who worked at the frontier of AI technology and infrastructure.
Be first to meet your companion.
Join the waitlist for the private pilot in Q1 2027. Early members help shape our models and form the first community circles.