Before Jeff McDonald built data platforms, he built stages. His college subject was theatre arts and stage design, a credential that appears to have wandered into the wrong biography until he explains it. A stage designer thinks in sequences and constraints. Something must happen first so that something else can happen next. Several kinds of specialist must produce one coherent experience. The audience should understand what it sees without having to inspect every rope, lamp, brace, and hurried cue behind the curtain.
McDonald now runs Kythera Labs, the Franklin, Tennessee company he co-founded in 2019. It organizes, standardizes, links, and analyzes complicated healthcare and life sciences data. His career change was real, but the habits survived it. He still asks what belongs in view, what must happen out of sight, and how the whole production can tell a clear story.
It is an unusually apt training for the data business. Everybody wants the revelation at center stage. Few people buy a ticket to watch identity resolution, metadata management, or a pipeline being rebuilt after a source changes format. Yet those are the cues that determine whether the grand finale is insight or merely a confident-looking mistake.
“In hindsight, it’s not that different from stage design: you’re constantly asking what belongs on the stage, what’s happening behind the scenes, how everything fits together, and how to make the story clear.”Jeff McDonald
The pickup truck period
Early in his career, McDonald worked at a technology rollup, learning from experienced colleagues and developing an enterprise strategy for technology and services aimed at healthcare organizations. Corporate changes prompted a move. He joined Evariant, then a healthcare startup, with responsibility for innovation. That job, he says, set the trajectory for what followed.
What followed was Expression Health, a company McDonald started to develop analytics for providers. Its origin story comes with the sort of physical detail software histories usually lack. The young business did not have enough money to purchase the technology it needed. Cloud computing offered no easy rescue; McDonald recalls that Microsoft Azure was not ready for this particular job in 2014. So the company bought 50 used servers from a Google data center.
McDonald loaded them into the back of his pickup truck. The machines were installed in vacant space behind a beauty parlor and became an early Hadoop deployment. Somewhere, amid the hum of recycled hardware and whatever appointments were taking place through the wall, a healthcare analytics product began to exist.
The contraption did not become sacred. It got the team through a critical moment, and then they replaced it with conventional infrastructure. That last step is what rescues the episode from startup folklore. Resourcefulness is not an obligation to remain rustic. A workaround does its job when it creates enough time, evidence, or revenue to reach the next version.
Expression Health later merged with two other companies to form Trilliant Health, where McDonald served as chief product innovation officer. By then, the technical backbone of healthcare analytics had become the recurring theme of his work. Dashboards changed. Companies changed. The difficult material underneath them remained stubbornly familiar.
A company built below the answer
After Trilliant, McDonald brought together people who had worked at Evariant and Expression Health. They founded Kythera Labs around a thesis both broad and specific: every healthcare analytics use case runs into a data technology problem first. Information arrives from different systems, under different definitions, with gaps, overlaps, and identities that do not always line up. More data can multiply the confusion.
Kythera's answer was Wayfinder, a platform designed to handle the full route from sourcing and processing through management, integration, and access. The company built it natively on the Databricks Lakehouse from the start. McDonald calls that a deliberate architectural bet, made before the market had offered comforting applause.
sources
& link
& trace
answer
The present scale gives the backstage metaphor industrial dimensions. McDonald wrote in June 2026 that Wayfinder processes about two billion medical transactions and three billion prescription transactions each year. The work resolves identities across fragmented sources and governs transformations from ingestion to usable output. His own verdict is accurate and faintly comic: it is “unglamorous work.” Five billion annual transactions can make unglamorous look rather busy.
The visible milestones accumulated. Kythera announced partnerships with Datavant and Databricks in 2021. It became a CMS Qualified Entity and made its first Inc. 5000 appearance in 2024 at No. 4,249. The company rose to No. 1,945 in 2025 and No. 1,870 in 2026. Databricks named it the 2026 ISV Innovation Built-On Partner of the Year.
McDonald rarely describes those results as a solo act. His public accounts keep returning to teamwork: engineering, product, data science, customer success, and subject expertise moving together. The team at Kythera contains working relationships carried across earlier companies. In a sector enamored of raw speed, this is a quieter advantage. People who know how their colleagues think spend less time translating every cue from scratch.
Questions before conclusions
The instinct predates Kythera. In 2015, McDonald co-authored an article in Healthcare Financial Management about using large datasets to build an integrated service line. The article described a practical sequence: understand market demand, guide people through integrated care, and improve engagement. Even then, data was presented as a means of coordinating decisions, not as decoration for a report.
His later writing sharpened the distinction between possessing information and understanding its limits. Claims, records, laboratories, registries, and other sources are produced for different reasons. They do not arrive “answer-ready,” one of Kythera's favored phrases. A dataset can be large and still be incomplete for a particular question. Context does not become optional merely because the file is impressive.
That restraint also appears in how McDonald talks about mistakes. In a 2026 leadership interview, he argued that mistakes are essential and perfection impossible. It is a useful admission from an executive whose company sells confidence in data. Confidence, in this telling, does not mean pretending error has been abolished. It means making the path visible enough to test, correct, and improve.
He applies the same logic to customers starting with AI. At a 2026 healthcare marketing event, his advice was not to let fear dictate the decision. Alignment comes first: identify a practical early win, bring the relevant stakeholders into the room, and understand what the organization is prepared to operate. Buying capability is easier than arranging responsibility. The technology may arrive in a contract; trust has to be rehearsed.
“People I work with will tell you I’m fond of saying ‘crawl, walk, run’ - though these days, with AI, it feels like we’re sprinting all the time.”Jeff McDonald
Trust has an operating system
The “crawl, walk, run” phrase contains McDonald's favored contradiction. He wants decisiveness under ambiguity, but he also believes in sequence. AI has increased the pressure to leap directly from possibility to production. His more recent writing argues for the missing middle: governance, lineage, auditability, and a clear account of who owns a decision when a system is wrong.
In July 2026, Kythera launched its Clinical Semantic Bridge MCP on the Databricks Marketplace. It translates ordinary clinical language into structured code sets used by data systems. The product is a translation layer, which fits a career spent connecting audiences to backstage machinery. A question asked in familiar language has to become computable without losing its meaning along the way.
McDonald's longer view is toward specialized AI agents operating inside governed workflows, coordinated through an orchestration layer. He does not frame the ambition as a faster version of the dashboard. He wants systems whose behavior can earn trust over time. The phrase he used when Kythera made the 2026 Inc. list is even cleaner: healthcare needs “answers organizations can defend.”
Defend is the key verb. It requires a trail. It assumes an intelligent skeptic will ask where a number came from, what it includes, what it excludes, which definition applies, and whether the result can be reproduced. A model may generate an answer in seconds. The right to trust that answer is accumulated before the prompt is typed.
McDonald's story makes the hidden layer visible without pretending it is glamorous. Stage design taught him that an audience need not see every beam to benefit from sound construction. Expression Health taught him that imperfect infrastructure can be useful if the team knows when to move on. The ventures between Evariant and Kythera taught him that a durable problem is worth following through several corporate lives.
The amusing thing about backstage work is that success makes it disappear. The set holds. The cue lands. The scene proceeds. Nobody applauds the bolt. McDonald has built a career around bolts, cues, and data plumbing anyway, because the show above them cannot be clearer than the structure below.