Data-science function from scratch
Built and lead a cross-functional team spanning ML research, embedded integration, and product software — research to shipped device feature, in-house.
Problem
Taking a medical-imaging capability from research to a shipped device feature touches ML research, embedded integration, and product software. When those disciplines live in separate silos, the handoffs are where the work goes to die — a model that the embedded team cannot deploy, a device app that the research team cannot instrument.
The need was for a single in-house function that could own the whole arc, from an idea on the bench to a feature running on the device in a clinician's hands, without throwing the work over a wall at each stage.
Approach
I built and lead that cross-functional team, spanning ML research, embedded integration, and product software. The point of keeping it cross-functional is that the same group that trains a model also owns getting it onto the compute module and into the app that operates the device — so the research is shaped by the deployment constraints from day one.
Alongside the technical direction, I own hiring and career development across ML, embedded, and product software. Growing the function from one to many meant setting the standards for how research becomes production and keeping that rigor intact as the team scaled.
What shipped
The function delivers research-to-shipped-feature in-house, as a Director-level responsibility rather than a hand-off between vendors or silos — the team that does the science also ships the device capability.
The result is an organization that can carry a capability across every layer it needs to cross, and grow without losing the discipline that made the first version trustworthy.