Premature infants in Stanford's neonatal intensive care unit are the first patients to receive intravenous nutrition guided by an artificial intelligence system, as a startup spun out of the university's medical school begins clinical testing of its decision-support tool.
Takeoff41, founded by Stanford postdoctoral researcher Joe Phongpreecha and professor Nima Aghaeepour, built TPN2.0 to address a persistent problem in neonatal care: total parenteral nutrition (TPN), the IV therapy given to infants who cannot feed by mouth, is complex to formulate and prone to error. TPN errors are among the most frequently cited medication mistakes in NICUs, according to the company.
The tool was trained on more than 140,000 TPN prescriptions from over 9,000 patients across two hospital systems. In a blind evaluation, physicians rated its recommendations higher than current standard practice. Retrospective analysis found that deviations from the system's recommendations correlated with 3.3 times higher rates of necrotizing enterocolitis, five times higher cholestasis, and five times higher mortality.
Most AI in healthcare is predictive; it tells you what might happen. TPN2.0 is prescriptive. It tells you what to do about it.
A 260-patient feasibility study is now underway at Stanford, with collaborations at UCLA, Brookdale University Hospital and Medical Center, and Virtua Health. The Clinical Research Forum named the underlying research a 2026 Top 10 Clinical Research Achievement Award finalist.
NICU clinicians face the hardest daily decisions, like calibrating oxygen so precisely that too much can cause blindness. Yet they have almost none of the AI support tools available in other specialties.
Takeoff41 closed a $4.3 million seed round in March 2024, co-led by Basis Set Ventures, with participation from Social Impact Capital and the Sarah Smith Fund. The company also holds an NIH SBIR Fast Track grant.
