YesPress Profile
Founder / Scientist / Professor
David
David
Sontag
He spent a decade teaching machines to read medicine. Then he built the company that does it for hospitals.
He spent a decade teaching machines to read medicine. Then he built the company that does it for hospitals.
In a collaboration with a cancer organization, Layer Health pulled structured, research-grade data out of dozens of brand-new patient charts in a few hours. The old way took more than a year.
That single number is the whole pitch. Somewhere in every hospital there is a trained professional sifting through hundreds of pages per patient, hunting for the one note that says whether a tumor spread or a drug was changed. It is slow, it is expensive, and two careful humans reading the same chart will not always agree. David Sontag built a company around the conviction that the messiest text in medicine is exactly the kind of problem large language models were made for.
He is the co-founder and CEO of Layer Health, and he is also a professor at MIT - on partial leave since 2025 to chase the thing his lab had been circling for years. The bet is plain: the algorithms exist, the data exists, and the missing piece is a layer of AI that clinicians actually trust sitting on top of the electronic health record.
AI - especially large language models - are uniquely suited to process and interpret the messy, unstructured data found in medical records.David Sontag
Sontag studied computer science at the University of California, Berkeley, then went to MIT for a PhD under machine-learning figure Tommi Jaakkola, working on the math of approximate inference in probabilistic models. In 2010 his thesis won the George M. Sprowls Award, MIT's prize for the best doctoral work in computer science that year. The early Sontag was a theorist - graphical models, optimization, the kind of papers that win best-paper awards at NIPS, EMNLP and UAI. He has all three.
After a year as a postdoc at Microsoft Research New England, he took a faculty job at NYU's Courant Institute from 2011 to 2016. Then MIT called him home in 2017, naming him the Hermann L. F. von Helmholtz Career Development Professor and giving him a foot in both the computer-science department and the medical-engineering institute.
That dual appointment is the tell. Plenty of machine-learning professors stay in the comfort of benchmarks. Sontag pointed his lab at hospitals - building models that make clinical predictions, methods for causal questions in high-dimensional data, and the scaffolding for a smarter electronic health record. Faculty awards from Google, Facebook and Adobe and an NSF CAREER Award followed.
The work kept bumping into the same wall. The most valuable information about a patient lives in free text - the notes, the narratives, the things no dropdown menu ever captured. You can build a brilliant model, but if it cannot read the chart, it is blind. Layer Health is what happened when Sontag decided to knock the wall down.
Layer Health reads both the structured fields and the unstructured notes in a patient's record, then abstracts them into the data registries, research datasets and care decisions that depend on getting it right.
Unlike rules-based software that fires on keywords, the platform reasons across a patient's entire chart - handling the nuanced, contradictory cases that break older tools.
It powers complex chart abstraction for cardiovascular, oncology and national surgery registries - the painstaking work hospitals usually staff with armies of reviewers.
The founding team pairs MIT ML researchers with a practicing emergency physician, so the output is meant to be trusted at the bedside, not just benchmarked.
Layer Health did not assemble a team of strangers. Sontag founded it alongside people he had already taught and worked with - Divya Gopinath and Luke Murray running engineering, Monica Agrawal leading research, and emergency physician Steven Horng on clinical informatics. The research group became the founding team.
The company is backed by Define Ventures, GV, Flare Capital Partners and MultiCare Capital Partners, and advised by health-tech names like John Halamka and David Axelrod. Early deployments at systems including Froedtert & the Medical College of Wisconsin, Intermountain Health and White Plains Hospital gave the pitch its proof points.
The vision for Layer Health is to become the foundational AI layer across healthcare organizations, enabling smarter decision-making at every level.David Sontag, on what he is building