You will own how messy, unstructured capture becomes understanding.

Work arrives from the field as raw video, audio, and images. Your job is to turn that into a grounded, timestamped model of what actually happened: what was seen, what condition it was in, what was done and when. Everything downstream is assembled from the building blocks you create.

This is applied work, not research. You will choose models, fuse them, measure them ruthlessly, ship them, and make them better every week with data coming back from the field. You will work directly with the founders, in person in New York City, and set the technical direction for the most important problem in the company.

If you want a clean, well-scoped problem with a settled roadmap, this is the wrong role. If you want to define how machines understand physical work, keep reading.