The clinical notes pile up faster than anyone can read them—physician observations, discharge summaries, imaging reports, all swimming in the unstructured chaos of modern healthcare documentation. Somewhere in that morass sits money, billions of it, tied to a coding system most patients have never heard of but one that determines how Medicare Advantage plans get paid.
Keebler Health thinks it's found a way through. The Durham-based startup announced a $16 million Series A round on April 15, 2026, led by Flare Capital Partners with backing from Sands Capital and a broad syndicate of returning investors. It's a substantial vote of confidence for a company barely three years old, now sitting on $23 million in total funding, all aimed at solving what CEO Isaac Park describes as healthcare's "approach problem" when it comes to risk adjustment.
The challenge is thornier than it sounds. Medicare's hierarchical condition category system—HCC codes, in the alphabet soup of healthcare acronyms—governs how the government pays managed care plans based on patient acuity. Sicker patients mean higher reimbursements, but only if those conditions are properly documented and captured. According to Keebler, missed chronic diagnoses buried in specialist notes represent revenue left on the table, or worse, audit liabilities waiting to surface.
Keebler's pitch centers on a transformer-based AI platform built specifically for this task. The company says its large language model can parse unstructured clinical data and extract relevant HCC codes while linking each finding back to source documentation—a feature designed with an eye toward the Risk Adjustment Data Validation audits that have become increasingly consequential under recent CMS policy shifts.
"They've built a platform that aligns with how clinical information is actually documented," said Ian Chiang, a partner at Flare Capital Partners who joined Keebler's board with the new round.
Perhaps the timing helps explain investor enthusiasm. CMS-HCC version 28 completed its three-year phase-in with 33% weighting for payment year 2024, 67% for payment year 2025, and reaching full 100% implementation for the 2026 payment year—forcing Medicare Advantage plans to overhaul their coding operations throughout the transition. And then there's the January 2023 final rule from the Centers for Medicare & Medicaid Services, which gave the agency authority to extrapolate RADV audit findings across an entire plan's book of business—turning what was once a compliance exercise into something with actual financial teeth.
The company positions itself as addressing both retrospective chart reviews and real-time concurrent workflows, though it's careful to note its system operates "without requiring clinician disruption." That's healthcare tech speak for: we don't add to the physician's administrative burden, already a flashpoint in an industry drowning in documentation requirements.
Keebler cites industry data suggesting roughly 80 percent of healthcare information remains unstructured, with the company's analysis indicating chronic condition capture consistency across different EHR sources at just 59.4 percent. Whether that's a data quality problem or a data accessibility problem depends on who you ask, but either way, it's inefficiency at scale.
A Familiar Fundraising Arc
The syndicate that came together for the Series A reads like a who's who of early-stage healthcare and enterprise investors, ten firms in all beyond the two leads. Freestyle Capital, which led Keebler's $6 million seed round announced in February 2025, returned for the new raise. So did New Stack Ventures, which anchored the company's $1.8 million pre-seed in August 2023. Also participating: Aviano Ventures, Everywhere Ventures, MBX Capital, Tau Ventures, Tweener Fund, Underdog Labs, and Hustle Fund.
That's a lot of checks for a Series A, which can signal either broad conviction or a round that came together in pieces—or both. The company isn't saying.
Park founded Keebler alongside Andrew Stickney (now COO) and Kevin Hill (CTO), later adding Terrell Bacchus as chief medical officer and Jeremy Powell as chief commercial officer. LinkedIn pegs the headcount somewhere between 11 and 50 employees as of mid-April, though anyone who's tracked startup hiring knows those public counts tend to lag reality by weeks or months.
The fresh capital will fund the usual suspects: team expansion, commercial scaling, infrastructure buildout. But Keebler also flagged plans to expand into RADV audit readiness workflows, a tacit acknowledgment that the regulatory environment has made audit defense as important as initial coding accuracy.
"Risk adjustment doesn't just have an awareness problem, it has an approach problem," Park said in the announcement—a line that doubles as both diagnosis and sales pitch.
Christy Steele, the Sands Capital partner who participated in the round, framed the opportunity around fragmented data. "Accurate risk adjustment depends on a complete and consistent view of the patient," she said. Fair enough, though one could say the same about nearly every problem in healthcare IT.
The Unstructured Data Bet
What Keebler is really betting on is that language models have gotten good enough to handle the messy reality of clinical documentation. Not the neat, structured fields of a lab result, but the narrative chaos of a physician's assessment or a radiologist's impression. The company says its platform integrates with FHIR standards and existing EHR systems, supporting CMS-HCC version 28 specifications out of the gate.
Whether that technical architecture translates into market traction remains to be seen. Keebler hasn't disclosed customer names or provided much detail on deployment scale, instead emphasizing its integrations with electronic health record systems and health information exchanges. For a company in growth mode, that reticence is notable—possibly a function of early traction still ramping, or contractual restrictions, or both.
The broader market for risk adjustment technology isn't new; legacy vendors and newer AI-focused startups have been circling this space for years. What's shifted is the regulatory pressure and the underlying capability of the models. A platform that can credibly automate HCC capture across fragmented data sources, maintain audit trails, and integrate into existing workflows without adding clinician burden? That would be valuable.
Whether Keebler has built it, and whether the market will pay for it at scale, is the $23 million question—literally.
For now, the company has runway and a roster of backers willing to find out.
