A clinically serious algorithm, treated like a side project
Steritas began with an expert algorithm for measuring glucocorticoid toxicity, developed by John Stone at Harvard University and Massachusetts General Hospital. There was already an app around it, and physicians were already using it.
The obvious problem was bugs. The deeper one was structural. The work had been treated like a side project, with a reactive team and a product model that asked one app to do too many jobs at once.
Physicians needed a tool they could keep using in practice. Steritas also needed a way to license the same algorithm to pharmaceutical customers who wanted to integrate it into studies. That meant serving two very different audiences with very different constraints.
“They helped us turn an early product into something a serious healthcare buyer could actually trust.”
Two implementations, one underlying algorithm
So the work was not just to build a better app. It was to give the algorithm the right delivery model. We rebuilt the physician-facing product so doctors could keep using it, while also creating a separate commercial path for research customers.
That second path took the form of a Python library, a licensing mechanism, and the technical protections needed to distribute the algorithm to paying customers in a more controlled way. The same underlying logic could now be used in practice by physicians and integrated into studies by pharmaceutical teams.
The physician-facing product could also keep generating the usage data Steritas relied on, while the licensed path gave paying customers a supported way to use the algorithm in research.
Around both implementations, we added publishing pipelines, automated tests, and regression tooling so algorithm issues could be diagnosed, fixed, and verified without guesswork.
The software was not the only thing that needed structure
Steritas also did not need another reactive provider. The company needed a partner who could support physicians and commercial research customers at the same time, while working with the compliance expectations that came with both.
The engagement changed over time. Early on, it looked more like a conventional project, with managers, engineers, and sprint planning. Once the core system was in place, the work shifted toward operations: SLA-backed support, diagnosis of algorithm issues, and regression work to keep changes controlled.
That left Steritas with a product system that matched the business more closely. Physicians could keep using the algorithm. Pharmaceutical customers had a supported licensing path. The company had a steadier way to maintain both. Steritas was lateracquired by IQVIA.
