Ubenwa

Market position

The company describes itself as the first to develop technology for interpreting infant cry sounds and as a first mover in what it calls a $200B pediatric care market, citing the world's largest clinically annotated infant cry database as a key asset. External coverage characterizes it as a leader in automated sound-based newborn diagnostics and as a Mila spinout backed by prominent AI investors.

Differentiators cited include a proprietary, clinically annotated infant cry database built with hospitals on three continents (compared with a prior public database of about 70 samples), a foundation model for infant cries, peer-reviewed and preprint research output with Mila-affiliated authors including Yoshua Bengio and Doina Precup, and a team combining AI researchers with neonatology expertise.

Technology

Ubenwa combines audio signal processing with machine learning, including neural networks trained on spectrograms and self-supervised learning, to extract acoustic biomarkers from voice and infant cries. The company describes having built the first foundation model for infant cries, with algorithms for cry activity tracking, biomarker detection, harmful-noise tracking, pain-cry identification and anomaly prediction that can run on any microphone-enabled device. Its dataset, described as the world's largest clinically annotated database of infant cry sounds, was assembled with pediatric hospitals across three continents; reported model performance from the clinical study averaged 89% sensitivity in identifying babies with health issues and 86% specificity in identifying healthy babies, and an early pilot on neurological injury from birth asphyxia showed roughly 40% improvement over APGAR scoring. A separate product line, Roseline AI, is an AI voice agent for clinic phone workflows.

Go-to-market

The company reaches consumers directly through the Nanni AI mobile app and offers demo bookings and a partner program on its website for healthcare organizations. Clinical credibility and data are built through research collaborations with children's hospitals and neonatologists across North America, South America and Africa, and its Roseline AI product is sold to healthcare providers, reported as deployed in more than 50 locations.

New parents; healthcare practitioners and clinics (including delivery rooms, neonatal and pediatric intensive care units and nurseries); academic researchers; and contract research organizations running clinical trials.

Geography

Headquartered in Montreal, Canada, with research and clinical partnerships spanning North America, South America and Africa, including Montreal Children's Hospital, Enugu State University Teaching Hospital and Rivers State Teaching Hospital in Nigeria, and Santa Casa de Misericordia in Brazil. The Nanni AI app is reported to have users in 200 countries, and Roseline AI is described as addressing healthcare access in Canada.

History

Onu's cry-based diagnosis research dates to at least 2014, with a 2017 workshop paper at NeurIPS presenting an SVM approach to cry-based diagnosis of birth asphyxia under the Ubenwa name, followed by neural transfer learning work at INTERSPEECH 2019 and self-supervised learning and dataset papers in 2023. As a Mila spinout, the company ran a three-year clinical study with five hospitals in Canada, Brazil and Nigeria, collecting roughly 10,000 clinically annotated cry samples, and open-sourced part of the dataset. In 2022 it announced a $2.5M pre-seed round led by Radical Ventures. It launched Nanni AI as a consumer app after demand from parents outpaced clinical deployment, then expanded into healthcare operations with the Roseline AI voice agent, reported in July 2025 as deployed in over 50 locations.

Risks & controversies

Coverage notes that the approach has met skepticism from outside observers, and that the company was continuing to gather data with doctors in Nigeria, Canada and Brazil to conduct further clinical validations. Its diagnostic ambitions depend on clinical study outcomes and on the procedures and approvals required for use in clinical practice, which the company cited as a reason it first launched a consumer app rather than a clinical deployment.