AI Support Pediatrics and Developmental Medicine.
PediaMed AI
Our main goal is to explore how AI agents can transform pediatric care. We are building a new breed of AI systems that (1) enable in-home, automatic screening for developmental disorders, (2) act as safe and engaging AI companions that grow up with children, and (3) push the boundaries of child-like continual learning and cognitive development in next-generation AI.
01 — In-Home Developmental Screening
Development is continuous, but clinical screening is episodic. Conditions like autism and cerebral palsy are most treatable when caught early — yet most children are diagnosed years after the first signs appear, and families in underserved regions may never be screened at all.
We build multimodal AI systems that observe how children move, look, gesture, and play in their natural home environment — turning everyday moments into early developmental signals, and bringing screening from the clinic into the living room.
02 — Safe AI Companions for Children
Children are not small adults, and AI built for adults is not built for children. Today's AI systems are trained on adult language, aligned to adult intent, and evaluated by adult benchmarks.
We design AI companions and child-specific large language models from the ground up: agents that understand children's intent, adapt their language and feedback to each child's developmental stage, and remain aligned with psychologists, pediatricians, and parents. Safety is not a guardrail we add at the end; it is the foundation we build on.
03 — Child-Like Learning in Machines
Children are the most remarkable learners we know. From sparse, noisy, multimodal experience, they acquire language, common sense, and social understanding — continually, without forgetting, and without terabytes of labeled data.
We study how the mechanisms of child cognitive development — curiosity, social learning, continual adaptation — can shape the next generation of AI systems, and we map the boundaries of what child-like learning in machines can achieve.