BREAKING Layer Health raises $21M Series A led by Define Ventures Intermountain Health deploying Layer AI across 33 hospitals Froedtert cuts chart abstraction time 65% Year-long data extraction now done in hours Named to 2026 NY Digital Health 100 Spun out of MIT · 200+ peer-reviewed papers BREAKING Layer Health raises $21M Series A led by Define Ventures Intermountain Health deploying Layer AI across 33 hospitals Froedtert cuts chart abstraction time 65% Year-long data extraction now done in hours Named to 2026 NY Digital Health 100 Spun out of MIT · 200+ peer-reviewed papers
Healthcare AI · Boston, USA

Layer Health

The company teaching software to read the entire medical chart - every note, every scan, every messy line of it - so the people who care for patients can stop doing it by hand.

Layer Health
EXHIBIT A. The Layer Health calling card. Behind the logo: large language models doing the reading nobody volunteers for.
Who they are now

A roomful of PhDs versus the world's most tedious paperwork

Somewhere in a hospital tonight, a trained clinician is squinting at a patient record, copying one data point into a registry form. Then the next. Then a few hundred more. This is medical chart review, and it is the unglamorous engine room of modern healthcare - quality reporting, research, billing, all of it depends on humans reading charts line by line.

Layer Health is the company betting that a large language model can do that reading - and do it at human level or better. Founded in 2023 and spun out of MIT, it sells an AI platform to hospitals and health systems that reasons over a patient's full longitudinal record and answers complex clinical questions, with the receipts to show its work. As of 2025 it had raised $25M, employed roughly 45 people, and signed up health systems most startups only put on a wish slide.

"The AI layer for chart review - to unlock value and improve care by identifying insights from longitudinal medical charts at scale."Layer Health, company tagline
The problem they saw

The chart is full of answers. Nobody has time to read it.

The electronic health record was supposed to make medicine legible. Instead it created an ocean. A single patient can generate thousands of pages across notes, labs, imaging reports and discharge summaries - structured fields wrapped around unstructured prose. The answers a hospital needs are all in there. Getting them out is the problem.

So health systems hire people to abstract charts by hand. It is slow, it is expensive, and it does not scale. A cancer organization once spent more than a year extracting real-world data for a research cohort. Quality teams spend weeks per registry. The irony is hard to miss: the more data healthcare collects, the harder it becomes to actually use any of it.

"Medical chart review has historically been costly and time-consuming - yet scaling it reduces friction across all of healthcare."David Sontag, CEO & Co-Founder (paraphrased)
1,000s
pages in a single chart
>1 yr
old way: one research cohort
Weeks
per registry, by hand
The founders' bet

Four researchers who'd already read the literature

Layer Health was not started by people learning healthcare on the fly. The founding team came out of MIT, Harvard Medical School, Microsoft and Google, carrying more than 200 peer-reviewed publications between them. CEO David Sontag is an MIT professor whose machine-learning research on clinical data is, in effect, the company's foundation.

Their bet was specific. Not that AI would replace clinicians - that it would do the reading clinicians never wanted to do, and prove every answer with a citation back to the source note. The trick was trust: a model that says "this patient had a stroke" is useless unless it can point to the line in the chart that says so. So that became the product's spine - evidence-based justifications, validated on each customer's own data before anything goes live.

David Sontag
CEO & Co-Founder · MIT
Monica Agrawal
Co-Founder
Luke Murray
Co-Founder
Divya Gopinath
Co-Founder
"Backed by GV, General Catalyst and Inception Health at seed - then Define Ventures at Series A. Investors who read footnotes."On the cap table
The short, busy history

From MIT lab to 33 hospitals in roughly two years

2023

Spun out of MIT

Layer Health launches with $4M in seed funding from GV (Google Ventures), General Catalyst and Inception Health.

2025 · MARCH

$21M Series A

Define Ventures leads the round, joined by Flare Capital Partners, GV and MultiCare Capital Partners, to scale AI-powered chart review.

2025 · JUNE

Intermountain Health deal

Strategic investment from Intermountain Ventures plus a multi-year deployment across stroke, bariatric and cardiovascular registries - spanning 33 hospitals.

2025

CB Insights recognition

Named one of the 50 Most Promising Digital Health Startups.

2026

NY Digital Health 100

Added to the 2026 New York Digital Health 100 - a who's-who of health tech worth watching.

The product

One AI layer, pointed at the work people dread

Layer Health's platform - the chart-reading engine its team calls Distill - sits on top of the EHR and turns structured and unstructured data into answers. It is built to drop into existing workflows rather than ask hospitals to rebuild them, and it keeps watching its own accuracy after deployment. Three jobs it does today:

01

Clinical Registry Automation

AI-supported abstraction for cardiovascular, surgical and oncology registries - 50%+ validated time savings in 10 weeks or less.

02

Clinical Pathways

Converts records into guideline-aligned care decisions and flags patients eligible for high-value procedures.

03

Site of Care Optimization

Redirects low-acuity cases to the right setting, freeing hospital capacity for complex procedures.

"It performs at human level or better - and it has to prove that on your data before it earns a seat."On how Layer Health validates accuracy
The proof

Numbers a skeptic can hold onto

A demo is easy. A health system rewriting a workflow is not. The reason Layer Health's customer list reads the way it does - Intermountain Health, MultiCare, Froedtert & Medical College of Wisconsin, White Plains Hospital - is that the time savings showed up in their own data, not a sales deck.

Chart abstraction time: before vs. with Layer Health

Lower is faster. Figures reflect reported customer results.
Manual baseline
100%
Froedtert quality abstraction
-65%
Registry abstraction (validated)
-50%+
Source: Layer Health customer reports (Froedtert; registry deployments in 10 weeks or less).

And the extreme case is the one that sticks: a leading cancer organization completed a real-world data extraction in hours that had previously taken more than a year. That is not a productivity bump. That is a different unit of time.

"Year to hours. The chart didn't get smaller - the reader got faster."On the cancer cohort extraction
The mission

Trustworthy answers, not just faster typing

Layer Health frames its work as unlocking value and improving care by reading charts at scale. The careful part is the word "trustworthy." In healthcare, a fast wrong answer is worse than a slow right one, which is why the platform is HIPAA-compliant, SOC 2 Type 2 certified, and validated on client data before deployment. The model earns its keep one verified citation at a time.

The business model follows the mission: B2B software sold to hospitals and health systems, deployed into the registry, quality and clinical-data-management teams who feel the chart-review burden most. Double their productivity, the pitch goes, and the savings pay for themselves.

"Responsible AI in healthcare isn't a slogan here - it's the only version that a hospital compliance officer will sign."On why validation comes first
Why it matters tomorrow

The reader nobody had to hire

Healthcare's data problem is not going to shrink. More notes, more scans, more registries, more regulatory reporting. Every trend points toward more reading, and there are not more clinicians to do it. If an AI layer can read reliably and prove its answers, it changes which questions a health system can even afford to ask - of its quality, its research, its operations.

Go back to that hospital tonight. The clinician squinting at a record, copying one data point into a form. In a building running Layer Health, that work is largely done - the model has read the chart, pulled the variable, and cited the note it came from. The clinician checks it instead of doing it. Same chart, same questions. Different reader, and a great deal more time for the patient in the next room.

$25M
total raised
~45
employees
2023
founded · MIT spinout
33
hospitals via Intermountain