Trustworthy AI for the smallest patients.
PediaMed AI is a pediatric AI research lab building interpretable vision and multimodal models for early detection and clinician-readable explanations.
Pediatric medicine deserves the same generation of AI tools as oncology and radiology — designed for the bodies, faces, and rare diseases of children.
We are building them with clinicians, for children.
- — Interpretable
Every prediction comes with a faithful, clinician-readable explanation. No black boxes at the bedside.
- — Pediatric-first
Models trained, validated, and benchmarked on children — not adapted from adult datasets after the fact.
- — Open by default
Datasets, weights, and methods released openly. Knowledge belongs to the field.
From question to release.
- — 01 / Define
A pediatric clinical question, scoped with hospital and university collaborators.
- — 02 / Build
Vision and multimodal models with reference implementations and training recipes.
- — 03 / Validate
Calibration error, robustness checks, and — where applicable — clinician review.
- — 04 / Release
Datasets, weights, and methods released openly. Knowledge belongs to the field.
Three themes, one method.
Clinician-Defensible AI
Calibrated, Lipschitz-bounded transformers that know what they don't know. Uncertainty isn't a feature — it's a clinical requirement.
Pediatric Behavioral & Vision Models
Vision transformers for facial and behavioral analysis — built to support, never replace, the clinical interview.
Open Tools for Underserved Children
Open datasets and infant pose estimation released to any lab that needs them. Impact measured in children reached, not papers cited.
Workshops & clinical training.

CVPR 2026 · Computer Vision for Children (CV4CHL)
Workshop on computer vision for children's development, health, and education. Hosting the Children's Gait Competition.

ICLR 2025 · AI for Children (AI4CHL)
Workshop bringing pediatricians, psychologists, educators, and AI researchers together. Recap and gallery inside.