Elliott Green

Elliott Green spends his days on a problem most people never see: the raw material of medicine is a mess, and the artificial intelligence meant to improve care is being built on only a sliver of it.

As co-founder and CEO of Dandelion Health, Green runs a New York company with a deceptively plain job description - assemble clean, diverse, real-world clinical data and hand it to the people building the next generation of medicine. The pitch sounds like plumbing. That is roughly the point. Green has built a career on the parts of healthcare that others find too tangled to touch, and he keeps coming back to the same conviction: better medicine starts with better data.

Dandelion, founded in 2020, is a real-world data and clinical AI platform. It pulls together electronic health records, medical images, waveforms such as ECGs, and unstructured clinical notes from a consortium of health systems, then packages the result as a clinical-grade dataset for life science companies and AI developers. In 2026 the company closed a $14 million Series A, capital earmarked for deeper data infrastructure, more pharma partnerships, and a larger commercial and scientific team.

The number Green returns to most is 80 percent. That is roughly the share of medical information locked inside images, waveforms, and notes - the unstructured material that rarely makes it into an AI training set. Structured lab values and billing codes are easy to query. Everything else, the richest signal a clinician actually reads, has been effectively invisible to most algorithms. Green's whole enterprise is a bet on surfacing that missing majority.

The 80 Percent Problem

To understand why Green cares so much about images and waveforms, look at where clinical AI usually gets its data. Most models learn from structured fields - the tidy, coded entries that hospitals bill against. Useful, but thin. The texture of a patient's story lives elsewhere: in the radiologist's read, the cardiologist's tracing, the note a physician typed at the end of a long shift.

Dandelion's answer is multimodal by design. By stitching together the structured and the unstructured, and by keeping records longitudinal - following a patient across systems and over time rather than freezing a single snapshot - the company aims to give developers something closer to clinical-trial quality without the cost and delay of running a trial from scratch.

Built Against Bias

Green is blunt about a flaw baked into decades of clinical research: it has skewed toward older, wealthier, whiter, and male populations. Algorithms trained on that history inherit its blind spots. Dandelion's response is structural rather than rhetorical. The company partners with regional, non-academic health systems - names like Sharp, Sanford, and Texas Health - specifically so that the data reflects a wider range of racial, ethnic, socioeconomic, and geographic realities.

That instinct is reinforced by his co-founders. Green built Dandelion with operator Niyum Gandhi and academics Sendhil Mullainathan and Ziad Obermeyer, the latter two known for influential work on algorithmic bias in healthcare. The pairing of hands-on operators with researchers who study how models go wrong is not accidental. It is the company's founding logic.

Dandelion has also launched what it describes as the first free validation service for healthcare AI safety and equity - a way for developers to test whether an algorithm holds up across different populations before it reaches patients. Giving that away for free is a statement of priorities as much as a product decision.

From Insurance to Infrastructure

Green did not start in health data. He studied economics and international relations at the University of Warwick, then went into finance, building the strategy and modeling muscles that show up later in how he frames problems. His pivot into healthcare came at Oscar Health, where he was a founding member as the company scrambled to respond to the newly passed Affordable Care Act.

At Oscar he helped expand the insurance business into ten additional states and built two notable partnerships - a joint venture with the Cleveland Clinic in Ohio and a tie-up with Humana in Nashville. That work put him at the seam between insurers, providers, and patients, which is exactly where healthcare's data problems tend to surface.

He kept moving down the stack. As head of commercial strategy and partnerships at TrialSpark, now Formation Bio, he worked on the clinical trials side. As SVP and general manager at Clarify Health, he focused on improving the patient journey through data. Each stop circled the same gap - fragmented, biased, hard-to-use information - until he decided to build the foundation himself.

The GLP-1 Bet

In 2024 Dandelion launched a GLP-1 Data Library, a multimodal real-world dataset built around the fast-growing class of drugs reshaping metabolic care. It draws on longitudinal records for millions of patients, roughly 200,000 of whom are on GLP-1 agonists, refreshed quarterly and paired with images, ECG waveforms, and notes.

Green's framing of it is characteristically about the gap he wants to close. Obesity care has advanced quickly, he argues, while the broader treatment of cardiometabolic disease still leans on one-size-fits-all approaches that immunology and oncology left behind years ago.

He has made a related argument about clinical trials themselves: that they run long, bloated, and expensive in part because the biological baselines used to design them lag behind the disease trajectories seen in real, modern patients. Fresh, longitudinal, real-world data is his proposed corrective - a way to design smarter studies from a truer starting point.

What He's After

Ask what Green is building toward and three words keep recurring: precision, personalization, and trust. He talks about an ecosystem where faster diagnoses and targeted therapies are accessible to all patients, not just those who happen to resemble the populations medicine has historically studied. It is an ambitious frame for a company whose day-to-day work is data pipelines and partnership contracts.

There is a consistency to the whole arc. Finance taught him to model. Oscar taught him to build at the messy intersections of the health system. TrialSpark and Clarify taught him where the data breaks. Dandelion is the attempt to fix the foundation so everything built on top of it can be better - and, he would insist, fairer.