The Company That Cleans Up Healthcare's Messiest Data Before Anyone Else Can Read It
Everybody wants to do AI on healthcare data. Almost nobody wants to spend two years cleaning it first. Kythera Labs turned that dirty job into a business.
There is a version of the healthcare-AI story you have heard a hundred times: a startup promises to predict readmissions, spot the next blockbuster drug market, or find the patients a hospital is missing. What that story skips is the part nobody films for the demo reel - the eighteen months of scrubbing before the model runs. American healthcare data arrives late, mislabeled, duplicated, and split across a dozen systems that do not agree on what a patient's name is. Kythera Labs, a company of fewer than forty people in Franklin, Tennessee, decided that mess was the product.
Founded in 2019, Kythera calls what it does remastering. The pitch is plain: hand over the raw, fragmented feed of claims and prescriptions, and get back data that is standardized, linked, enriched, de-identified, and - the phrase everyone wants now - AI-ready. The company's tagline, "Clearer answers from complex healthcare data," is less slogan than job description.
01 · The ScaleWhat a data company at national scale actually touches
The numbers are the easiest way to understand why this is hard. Each year Kythera processes roughly 2 billion medical transactions and about 3 billion prescription transactions, reflecting more than 300 million US patients, sitting on something close to three petabytes of data. That is not a dataset you open in a spreadsheet. It is closer to a small national utility.
02 · The ProductWayfinder, the platform built on a bet
Kythera's engine is a platform called Wayfinder - a cloud-based, machine-learning data science environment purpose-built for healthcare. It handles sourcing, correction, linking, and enrichment, then hands analysts working surfaces (branded WorkBoard and WorkSpace) and flexible deployment, whether a customer wants it in the cloud or behind their own firewall. The interesting part is what sits underneath: Kythera built the whole thing on Databricks, leaning on Spark Declarative Pipelines, Delta Sharing, and Unity Catalog.
CEO Jeff McDonald is blunt about how much that choice mattered. In Kythera's own telling, the platform cut query pipeline design from a job that took competitors up to two weeks down to a few minutes, and shortened time to market by roughly 80 percent.
03 · The MoatA certification you cannot buy with a funding round
Speed wins deals. Permission keeps them. In August 2024, Kythera was certified by the Centers for Medicare & Medicaid Services as a Qualified Entity - a designation that grants access to nationwide Medicare claims data across Parts A, B, and D, covering more than 66 million beneficiaries. Only a small group of organizations hold that status, and fewer still have nationwide reach.
This is the quiet strategic point. In a regulated market, the hardest thing for a competitor to copy is not your model - it is your right to touch the data in the first place. A rival can raise more money and hire better engineers and still not be allowed in the room.
04 · The CustomersSelling the plumbing, not the faucet
Kythera's customers are the companies you might otherwise assume built this themselves: hospitals and health systems, pharmaceutical and life sciences firms, insurers and reinsurers, payers, government entities, and - tellingly - other analytics vendors. In June 2025, Kythera announced partnerships with Preverity, an insurtech analytics firm focused on patient safety and malpractice risk, and GAM, which works on healthcare waste reduction and care efficiency. In both cases Kythera is the data infrastructure underneath; the partners put their brand on the insights.
That is the whole business model in one sentence. Kythera is a picks-and-shovels play for healthcare AI: it would rather be the layer every insight is built on than fight for the insight itself. The economics only work because it is infrastructure - a team of roughly 38 people, on an estimated $2.8 million in revenue, operating at a scale of 300 million patients. You cannot get that ratio selling consulting hours. You get it by cleaning the data once and licensing the result to everyone.
05 · The FieldWhere it sits, and who it is up against
Kythera is not alone in the real-world healthcare data market. It competes, depending on the deal, with names like Tuva Health, Datavant, HealthVerity, and Komodo Health - and, just as often, with a customer's own internal data-engineering team that has decided to build claims pipelines from scratch. Kythera's counter-pitch leans on quality over bulk: it argues for "real patient activity over dataset volume alone," a jab at bigger rivals that brag about size. Whether that holds as the market consolidates is the open question. For now, being small, fast, and CMS-certified has been enough to land on the 2025 Inc. 5000.
06 · The PeopleTwenty years of healthcare, on purpose
The founding team is not a group of outsiders who discovered healthcare had a data problem. Jeff McDonald spent more than two decades in the industry before starting the company. Co-founders Ryan Leurck (now Chief Analytics Officer, a former Georgia Tech aerospace systems researcher), Matt Ryan (CTO, a recognized Databricks innovator), and Russ Sacks (EVP of Innovation and Data Science, a Lean Six Sigma Master Black Belt) round out a bench where most executives bring 20-plus years each. The knowledge that everyone re-cleans the same data - and that nobody enjoys it - is the kind of insight you only get from having done it yourself.
The takeaway for anyone building in a hyped space is uncomfortable and useful: the unglamorous groundwork can be the whole company. While the market argues about which model is smartest, Kythera bet that the real bottleneck sits one layer down, in the data that model has to trust. So far, the bet is paying rent.