* DraftAid is building the intelligence layer for mechanical engineering. We started by auto-generating manufacturing drawings from 3D CAD parts. We are now building representations that enable us to go much further.
### What you'll do
* Design learned representations over a large corpus of 3D assemblies and their associated manufacturing drawings
* Train and evaluate models that drive drawing generation decisions
* Build the data and training infrastructure from scratch: pipelines, eval harnesses, dataset curation
* Integrate models into a production geometry engine written in C#
* Own the full ML stack. There is no existing ML team; you are it
* Own problems, not tickets
### What we're looking for
* Deep experience training encoder-decoder architectures and representation learning systems from scratch
* Practical experience building with LLMs as components in larger systems
* Comfort working with 3D data: meshes, B-rep, point clouds, or similar geometric representations
* The ability to look at a messy, domain-specific corpus and figure out what signal is in it
### Nice to have
* Experience with 3D world models and spatial reasoning systems
* Background in robotics perception, 3D reconstruction, NeRFs, or geometric deep learning
* Familiarity with C# or TypeScript
### What we offer
* Flexible hours and hybrid in-office
* Competitive salary and equity package.
* Small team, high ownership
Visa sponsorship is offered for this role.
