Dermatology AI that works for everyone.

Tona is closing the skin tone gap in medical AI — starting from the data, building diagnostic tools that perform equitably across all Fitzpatrick skin types.

<5%
of dermatology dataset images represent dark skin tones (Fitzpatrick V–VI)
3B+
people globally affected by biased dermatology AI
↑ Miss rate
AI diagnostic accuracy drops significantly on darker skin tones, leading to misdiagnosis and delayed treatment

Dermatology AI fails on dark skin tones.

The data gap is not a minor oversight — it is a structural flaw that causes real harm. When AI models are trained almost entirely on light skin tone images, darker skin patients receive worse diagnostic care.


Recent studies confirm: a 2025 review in Cureus found AI is significantly less effective for Fitzpatrick IV–VI patients. A 2024 review found only 30% of dermatology AI programs reported any data on skin of color.

<5%
Dark skin tone representation in popular dermatology training datasets
30%
Of dermatology AI programs report any data on skin of color usage
3B+
People globally with Fitzpatrick V–VI skin tones underserved by current AI


"Misdiagnosis isn't a bug — it's a consequence of biased training data. We fix the data first."

Fairness starts with data.

We're building AI that works equitably across all skin tones. Our approach is rooted in one principle: representative data produces representative results. By centering dark skin tones in our training process, we're fixing the bias at its source.

📊

Balanced Data

We prioritize dark skin tone representation from the ground up, ensuring every model learns from diverse, high-quality clinical examples.

🔬

Clinical Rigor

Our work is validated by dermatologists and grounded in peer-reviewed research. We're not cutting corners on accuracy.

🔒

Privacy First

Patients don't need to share their data to benefit from better AI. We respect privacy while closing the representation gap.

🎯

Real-World Results

Our goal is clinical adoption. Better diagnostic accuracy translates to better patient outcomes across all communities.

Where we are today.

Building from the ground up. Currently halfway through the data pipeline, with a working skin tone classifier already deployed.

✓
Data Collection
✓
SkinToneNet Classifier
⋯
Data Pipeline ← in progress
Augmentation Pipeline
Diagnostic Model
RAG + Vector DB

Built by someone who's always built things.

From Rwanda to Rhode Island — a founder with deep technical roots and a lifelong commitment to building for underserved communities.

Delyce
Founder & CEO
  • MSCS, Brown University — AI & Machine Learning
  • Robotics & diffusion policy research (steering generalization in unseen environments)
  • 1.5 years open source engineering — EnAccess (energy infrastructure across Africa)

Looking for co-founders

Clinician / Dermatologist

Medical validation, clinical workflow expertise, and help bridging the technical and clinical sides of Tona.

Business Co-founder

Go-to-market, sales, and operations. Experience in healthtech or enterprise SaaS preferred.


Reach out →

Interested in Tona?

Whether you're a potential co-founder, a dermatologist interested in collaborating, or an investor — we'd love to hear from you.

Send Email →