Cielo was inspired by the fragile first days of a newborn’s life, when subtle signs like slight breathing irregularities, faint changes in skin tone, or unusual cry patterns can signal serious conditions. Globally, 2.3 million newborns die each year within their first 28 days, and the neonatal period accounts for 47% of all deaths in children under five. Two major causes are birth asphyxia, responsible for about 900,000 deaths annually, and neonatal jaundice, which affects 60% of full-term and 80% of preterm babies, with severe cases causing over 100,000 deaths each year. Many of these deaths are preventable with early detection, yet continuous monitoring systems are expensive, scarce, or unavailable in many hospitals and homes.
Cielo, the Smart Early Observation System, is a multimodal AI platform that continuously observes newborns using computer vision, audio intelligence, and genomic insights. The video engine analyzes chest movement, breathing patterns, posture, and skin tone to detect early signs of conditions like jaundice and respiratory distress. At the same time, the audio engine listens to cry signals and analyzes their acoustic patterns to identify signs of pain, hunger, or neurological distress. To make the system more predictive and personalized, I added a genomics-driven module that analyzes genetic markers, predicts disease susceptibility, and estimates how a newborn might respond to certain medications. This allows the system to support safer treatment decisions and move toward more personalized neonatal care.
I built Cielo as a real-time multimodal pipeline that combines a video analysis system for breathing and skin tone detection, an audio pipeline for cry classification, and a genomics module for risk scoring and drug–gene interactions. An AI agent layer brings together signals from all modules and generates a simple, structured risk report categorized as Normal, Caution, or Critical. Through this process, I learned the importance of combining multiple biological signals, the complexity of neonatal behavior, and the challenges of designing systems that work in real time. One major challenge was signal variability, since newborn movements are unpredictable and required careful filtering. Cry classification was also difficult because different cry types often sound similar. Running video and audio models simultaneously required optimization for speed, and integrating the genomics component meant translating complex biological data into clear, understandable insights.
Cielo aims to make early neonatal monitoring accessible, intelligent, and continuous. By turning everyday devices into AI-powered observation systems, it has the potential to reduce preventable deaths, support hospitals with limited resources, and bring more proactive, personalized care to newborns around the world.
Log in or sign up for Devpost to join the conversation.