Epsilon Health Raises $20M. The Bigger Bet Is an AI-Native Radiology Practice, Not Another Radiology Tool.

Epsilon Health emerged from stealth with a $20 million Series A to scale a vertically integrated radiology practice built around physician radiologists, proprietary AI and production workflow infrastructure.

Epsilon Health has emerged from stealth with a $20 million Series A, entering one of healthcare AI's most mature categories with a model that looks very different from another point solution.

Epsilon describes itself as an AI-native radiology practice. The company combines physician radiologists with proprietary AI and a technology stack designed around the practice itself rather than selling standalone software into someone else's workflow.

Radiology AI has had a deployment problem

Radiology has been an AI proving ground for more than a decade. Yet the core capacity problem has not disappeared.

Epsilon says U.S. imaging volume is approaching 700 million scans annually while the radiologist workforce is shrinking. The company argues that point solutions, fragmented integrations and models that fail to generalize across sites have prevented AI from meaningfully closing that gap at scale.

That diagnosis mirrors a broader pattern RCAI is tracking across the RCAI Market Map: AI creates more value when it owns a meaningful portion of the workflow instead of sitting beside it.

The practice is the product

Epsilon's architecture changes the commercial model.

Traditional radiology AI vendors typically sell detection, prioritization or workflow software to imaging organizations. Epsilon can instead deploy its technology directly inside its own clinical operating model, iterate against real production data and capture the economic value of higher radiologist throughput itself.

The company says its model brings radiologists and technologists into one operating system and allows it to deploy and improve AI without the integration barriers that often slow third-party software adoption.

That makes Epsilon closer to an AI-enabled services company with proprietary infrastructure than conventional healthcare SaaS.

This is the same vertical-integration thesis appearing elsewhere in healthcare AI

Healthcare AI companies are increasingly moving beyond recommendation software toward ownership of execution.

In revenue cycle, GenHealth.ai is positioning agents as digital labor that completes administrative work. Stedi is building infrastructure underneath increasingly autonomous revenue-cycle workflows. And ARPA-H is now funding a regulated architecture for autonomous clinical AI that can act, be supervised and escalate exceptions.

Epsilon pushes the idea one step further: combine the AI with the organization that actually delivers the service.

Why the model could compound

A vertically integrated practice potentially creates a feedback loop that standalone software vendors struggle to reproduce.

More studies create more production data. More production data can improve models and workflows. Better workflows can increase radiologist capacity. Higher capacity can support more imaging volume, which creates still more data.

That doesn't guarantee the model works. Radiology remains a regulated, quality-sensitive clinical service with difficult recruiting and operational requirements.

But if Epsilon can demonstrate both clinical quality and materially better unit economics, it could turn the radiologist shortage from a constraint into the reason customers adopt the model.

The Series A is a bet on infrastructure, not a feature

Axios reported that Epsilon emerged from stealth with the $20 million Series A. The company is led by founder and CEO Rustin Rassoli and says its team includes experience from AI, technology and radiology organizations.

The capital is notable because imaging already has a large field of AI vendors. Investors are not underwriting an untouched category.

They are underwriting a different answer to the question of how AI gets deployed.

That is increasingly the question that matters across healthcare.

RCAI Take

The radiology market has spent years asking whether AI can read images.

Epsilon is asking a more commercially important question: what if you redesign the radiology practice around AI from the beginning?

That moves the competitive battleground away from model accuracy alone and toward workflow ownership, clinician productivity, proprietary data, operating execution and ultimately the cost and speed of delivering a final interpretation.

It is the same shift happening across administrative AI: the value migrates from recommending work to actually doing it.

If Epsilon succeeds, the lesson will extend well beyond radiology. The next generation of healthcare AI companies may increasingly look less like software vendors and more like technology-native operators.

Track Epsilon and other clinical and administrative AI companies in the RCAI Market Map, and follow financings in the Deals & Raises tracker.