NinaOS — Personal AI Agent for Primary Infancy
🚀 Elevator Pitch
A personal AI doctor for the most critical days of life. NinaOS is a personal health agent for primary infancy, starting with early detection of biliary atresia.
Inspiration
NinaOS was born from a structural failure in early childhood healthcare.
Biliary atresia is the leading cause of pediatric liver transplants worldwide. When diagnosis is late, survival without transplant is almost zero.
What makes this unacceptable is that early signs appear at home, every day — yet detection still depends on memory, chance, and delayed medical visits.
This is not only a biliary atresia problem. It is a primary infancy problem.
The first months of life are the most fragile, the most time-sensitive, and the least continuously observed.
We chose biliary atresia as our starting point because it perfectly exposes this gap. But our real ambition is bigger:
to build the personal AI agent layer for primary infancy.
A system that observes continuously, remembers over time, and escalates risk early — starting with this case.
What it does
NinaOS is a personal AI doctor for primary infancy.
It acts as a continuous health agent for newborns, starting with early detection of biliary atresia.
NinaOS:
- Analyzes diaper images to detect abnormal stool color patterns
- Tracks early neonatal risk signals
- Builds a longitudinal health memory for each baby
- Detects changes and emerging patterns over time
- Generates alerts when risk appears
- Produces structured summaries for pediatricians
Biliary atresia is our first clinical module. NinaOS is designed as a general agent framework for early childhood, able to expand to other conditions where time, patterns, and continuous vigilance are critical.
How we built it
We built NinaOS as an agent system, not a feature app.
The entire architecture follows one principle:
every baby should have a personal health agent.
Our system includes:
- An onboarding flow that initializes Nina as the baby’s personal AI doctor
- An image ingestion and processing pipeline
- A computer vision + reasoning layer
- A longitudinal memory system
- A risk and escalation engine
- A reporting layer to support pediatric conversations
From day one, the architecture was designed to support multiple early-infancy conditions, with biliary atresia as the first real-world case.
Challenges we ran into
- Avoiding a “single-disease app” trap
- Designing trust for an AI doctor in early childhood
- Translating lethal urgency into usable UX
- Defining what should be automated vs. escalated
- Structuring long-term memory that is clinically meaningful
The hardest challenge was not technical. It was conceptual: building something that behaves like an agent — not a gadget.
Accomplishments that we’re proud of
- Framing primary infancy as a personal-agent problem
- Designing an extensible agent architecture
- Delivering a working end-to-end prototype
- Creating a true agent activation onboarding
- Building a system that reasons over time, not isolated inputs
Most importantly, we are proud of not building “a baby app,” but the foundation of a personal AI agent for early life.
What we learned
- The biggest gap in early childhood care is continuous observation
- The real power of agents is memory + time, not isolated predictions
- Parents don’t need more content — they need delegated vigilance
- Starting with a lethal, time-critical case creates real product discipline
We didn’t just learn how to detect biliary atresia. We learned how to design an agent for primary infancy.
What’s next for NinaOS
Biliary atresia is our entry point — not our destination.
Next steps:
- Clinical validation with pediatric specialists
- Expansion to additional neonatal conditions
- Pediatric and hospital integrations
- Longitudinal risk modeling
- Supervised pilots with families
Our long-term goal is clear:
to build the personal AI agent layer for primary infancy — starting with this case and evolving into a full early-life health system.
Log in or sign up for Devpost to join the conversation.