Keebler Health Raises $16M as LLM-Native Risk Adjustment Challenges Legacy NLP

Keebler Health, a risk adjustment platform built natively on large language models, has closed a $16 million Series A led by Flare Capital Partners, with participation from Sands Capital and a syndicate of venture funds. The round brings total capital raised to $23 million since the company's 2023 founding.

Unlike legacy platforms that retrofit natural language processing to scan for keywords, Keebler processes raw clinical narrative to surface Hierarchical Condition Category coding opportunities used in Medicare Advantage reimbursement. The company plans to expand into compliance and audit workflows, specifically targeting AI-enabled Risk Adjustment Data Validation audit readiness as Centers for Medicare & Medicaid Services oversight tightens.

The funding arrives as healthcare payers face mounting pressure to extract accurate diagnostic codes from unstructured clinical documentation. Medicare Advantage plans rely on HCC codes to determine risk-adjusted payments, making coding accuracy a direct revenue driver. Regulatory scrutiny has escalated following CMS audits that identified billions in improper payments tied to unsupported diagnoses.

(Keebler's LLM-native architecture represents a departure from the retrofitted NLP tools that have dominated risk adjustment software for the past decade, many of which struggle with clinical nuance and context-dependent language.)

The healthcare AI market has seen parallel investment in adjacent infrastructure. InsightFinder, an AI observability platform founded by North Carolina State University computer science professor Helen Gu, recently raised $15 million in a Series B led by Yu Galaxy to address reliability issues in AI-driven enterprise systems. InsightFinder's Autonomous Reliability Insights product uses unsupervised machine learning and proprietary language models to diagnose failures across AI-integrated tech stacks, a capability Gu says extends beyond development-phase evaluation to production monitoring.

Meanwhile, Anthropic released Claude Opus 4.7, introducing multi-agent coordination and extended workflow focus designed to sustain coherence over hours-long agentic tasks. The company publicly acknowledged that Opus 4.7 does not match the performance of Mythos, an unreleased system withheld due to safety concerns. Forbes reported that the White House is considering granting government agencies access to Mythos, though no timeline has been disclosed.

Factory, an AI coding startup, is in talks to raise $150 million at a $1.5 billion valuation led by Khosla Ventures, with participation from Sequoia Capital, Insight Partners, and Blackstone. Venture capitalist Keith Rabois will join the board. The round underscores continued investor appetite for autonomous coding agents despite a crowded competitive landscape.

The healthcare vertical has emerged as a proving ground for LLM applications where accuracy and auditability carry regulatory and financial consequences. Risk adjustment represents a high-stakes use case: errors can trigger compliance penalties, while missed codes leave reimbursement on the table. Keebler's bet is that models trained to understand clinical context will outperform keyword-matching systems as documentation complexity increases and regulatory expectations evolve.