Founding Machine Learning Engineer

Colare's assessments are how leading space and defense, manufacturing, and robotics companies decide who their next engineers are. Automating the scoring behind those assessments is an open machine learning problem. You would own it.

The deterministic part is done. If a finished CAD model meets spec, Colare can already tell. What is not solved is scoring the process: how an engineer got there, what they considered, what they discarded. That means turning senior engineers' judgment into machine-readable rubrics, which is a different problem from the human checklist version of the same thing.

You would report to Nain, the cofounder and CTO, and work directly with the mechanical and electrical engineers who define what correct looks like.

What you'll do

Build the scoring models. Multimodal architectures reasoning over PCB layouts, CAD geometry, simulation output, and candidate interaction traces.

Encode the process, not just the result. Translate how good engineers work into rubrics a model can apply.

Own the learning loop. Data generation from real assessment runs, trajectory capture, failure analysis, dataset curation, continuous evaluation.

Work with the domain experts. Sit with the ME and EE to turn their judgment into structured labels.

Ship to production. Models go into scoring pipelines used in live hiring decisions.