Our mission: Democratizing medical diagnostics

WELES Open is a global, independent, non-profit initiative building an open AI ecosystem for ultrasound. We bridge medical science and software engineering, so a small clinic and a university hospital have access to the same quality of decision-support tools.

WHY “WELES”

An old name for an old idea: make the invisible visible

WELES stands for “Wide Evidence-based Learning Ecosystem Sonography.” The name also nods to Veles, the old Slavic guardian of wisdom and hidden things — fitting, since we use AI to help physicians catch the abnormalities that are hardest to see.

The three pillars

Evidence-based medicine

Models are designed, trained and validated by practicing physicians against rigorous clinical standards.

Pure open source

No black-box AI in medicine. Full source on GitHub — anyone can verify how it works.

Non-profit, by design

Funds and grants go straight into infrastructure, legal anonymized datasets, and development.

Under the hood

Two-stage independent review

Every image gets a first clinical read AND an independent second review — even a reported 'no pathology' is checked twice before it enters the dataset.

YOLO-based detection

Organ- and pathology-specific object detection models, trained and continuously re-validated on the growing, verified dataset.

Device-aware calibration

Each ultrasound machine model has its own profile, so the system knows exactly what part of the screen to read (and what to redact).

PRIVACY & DATA

Patient privacy is handled before data ever leaves the device profile.

Ultrasound machines often display patient-identifying information directly on screen. Because every device model has its own calibrated profile, the system knows exactly which regions of the image to redact before it's ever stored or reviewed — an automatic step, not something left to whoever uploads the image to remember.

Only anonymized, clinically-verified images become part of the shared dataset used to train and validate our models.

PATIENT: KOWALSKI J. ID: 88041207312 2026-07-16 10:24 anonymized

What we're building toward

Catch what's easy to miss

Detect abnormalities in ultrasound exams, even the subtle ones.

Support fast decisions

Help physicians and paramedics assess patients quickly, e.g. via FAST and POCUS protocols.

Automate the repetitive

Automate routine measurements to cut examination time.

Join the revolution. Let's build this together.

Software developer, AI engineer, doctor, or researcher — there's a place for you here.

Join us