Epsilon Health Emerges With $27.6M for AI-Native Radiology

The radiology AI market has spent years asking practices to install one more tool. Epsilon Health has raised $27.6M around a more operational bet: redesign the practice itself, put AI inside the daily reading workflow, and keep physician radiologists responsible for the final interpretation.

AlleyCorp led the financing, with Uncork Capital, Renegade Partners, SemperVirens, and Jack Altman participating. Epsilon emerged from stealth on September 10, 2026, with a San Francisco team spanning clinical radiology, machine learning, and production software.

The timing is bigger than another model announcement. The American College of Radiology says imaging volume continues to grow faster than the radiologist workforce, while practices face heavier workloads, consolidation, and burnout. Epsilon is trying to turn that capacity problem into the architecture of the company.

What Epsilon Health Announced

Epsilon's September 10 announcement says the company raised $27.6M in funding led by AlleyCorp. Uncork Capital, Renegade Partners, SemperVirens, and Jack Altman also participated; the announcement says Altman invested while serving as Managing Partner at Alt Capital.

The public capital record needs careful reading. Epsilon's homepage labels the announcement $27.6M in Series A Funding, while Axios reported a $20M Series A one day earlier. Renegade's portfolio separately labels its Epsilon position as a led seed investment. None of the public materials reviewed itemizes the difference, so the responsible conclusion is that Epsilon is now announcing $27.6M under a Series A banner, not that outsiders can reconstruct a $7.6M prior tranche from subtraction.

Epsilon did not disclose a valuation, ownership terms, revenue, or a round-by-round funding history. The company says the new capital will support market expansion, hiring, expanded clinical partnerships, and infrastructure investment.

The Practice Is the Product

Most radiology AI vendors sell software into a practice that already has a reading queue, staffing model, technology stack, customer commitments, and clinical accountability. Epsilon has chosen to become the practice. Its purpose-built AI sits inside the workflow of board-certified radiologists, while physicians review, edit, sign, and own the final report.

That structure changes where the value and the risk live. Epsilon can use production cases and radiologist corrections to improve models and workflows, potentially avoiding the familiar gap between a promising benchmark and a tool that clinicians struggle to adopt. The same integration also makes Epsilon responsible for recruiting radiologists, managing quality, connecting to imaging-center systems, protecting patient data, and delivering reports when customers expect them.

Uncork Capital's investment note says Epsilon's scans are interpreted 2-3x faster than before and describes the workflow as a capacity multiplier. That is an investor-reported claim, not an independently audited outcome. The more durable question is whether the practice can sustain faster interpretation while maintaining clinical quality across sites, modalities, patient populations, and rising volume.

Early Scale Meets a National Capacity Problem

Epsilon says it has served more than 250,000 patients and processes more than 2,500 imaging studies per day. The company also says it is on track to interpret roughly 1% of all daily U.S. X-rays in 2026 and already handles more than half the imaging volume for one unnamed large outpatient imaging provider. Those figures are company-reported, but they show why investors are treating the business as an operating system rather than a laboratory project.

Independent workforce evidence supports the pressure underneath the pitch. The ACR's 2026 workforce update describes imaging volumes growing faster than the number of radiologists and radiologists leaving practice at more than twice the rate seen less than a decade ago. RSNA reporting similarly points to an aging population, rising exam complexity, early retirements, and a training pipeline that cannot expand quickly.

The technology market is busy, too. The FDA's list of AI-enabled medical devices contains a long catalog of radiology products. The existence of many cleared tools has not removed the organizational work of selecting, integrating, monitoring, and using them inside a clinical service. Epsilon's wager is that the integration layer becomes more defensible when the technology company also carries the service obligation.

The Team Connects Clinical Work and Machine Learning

Founder and CEO Rustin Rassoli grew up around his father's imaging practice and has described seeing the patient consequences of missed or delayed diagnoses. The company story begins with that human pressure, then moves into a structural decision: build the radiology practice around AI instead of asking an outside practice to absorb another product.

Chief Medical Officer Dr. Roi Bittane previously served as CMO and Chief of Strategy at Envision Radiology, according to Epsilon's team page. Head of Machine Learning Arjun Karpur brings prior research experience at DeepMind and Google Research. Epsilon's public team also includes research engineer Aakash Mishra and software engineer Matt Naing.

AlleyCorp's lead investment fits that operating mix. Dr. Alexi Nazem, the firm's healthcare General Partner, framed Epsilon as a way to extend scarce diagnostic expertise. The investor thesis depends on physicians and technologists improving one system together, with each production case becoming operational feedback rather than another disconnected data point.

What the $27.6M Has to Prove

Epsilon has already moved beyond the easiest version of radiology AI: demonstrate a model, announce accuracy, and let the customer solve deployment. Owning the practice creates a tighter feedback loop and a clearer path to capturing the economics of productivity. It also removes the excuse that adoption failed somewhere outside the product.

The next evidence must therefore go beyond study volume. Healthcare operators will want to see quality controls, turnaround times, radiologist productivity, exception handling, integration reliability, customer retention, and performance across different clinical settings. Investors will eventually need proof that the service model can scale without making labor, infrastructure, or compliance costs rise just as quickly.

Epsilon's opportunity sits inside a difficult handoff: more images are being created, fewer specialists are available to interpret them, and patients still need a responsible physician to sign the answer. The $27.6M gives the company room to expand the practice, deepen the system, and show whether AI can move radiologist judgment through that handoff without weakening the trust attached to the final report.