Founding Engineer, 3D Machine Learning and Research

Job Description

MimoLabs builds Mimo, an agentic system for making physical things. Describe anything and Mimo helps you make it: he designs the geometry, simulates that it can be made, configures material, process and price, and routes it to production, across 3D printing, CNC, laser cutting and more. Mimo ships today as Mimo Studio for anyone regardless of skill, and we are building developer tools and enterprise solutions for more technical applications.

\n

\n

We are a VC funded, experienced technical team, based at Newlab in Detroit and remote across the US.

\n

\n

THE PROBLEM YOU WOULD OWN

\n

\n

Mimo generates manufacturable geometry and other assets with several methodologies. You will own the pipeline and tooling behind all of that. You will also lead the research practice for which paths we build next.

\n

\n

- Generation. The models and pipeline that turn intent into manufacturable assets, served fast and at scale.

\n

- Understanding. Part segmentation, semantic classification, and spatial comprehension.

\n

- Simulation. Verification and physics driven loops.

\n

- Training. Pre- and post-training pipelines and notebooks, checkpoints, and evaluations.

\n

\n

WHAT WE LOOK FOR

\n

\n

- Shipped or published 3D machine learning. Generation (diffusion or flow models over voxel, SDF or latent-set representations), segmentation, or reconstruction, with results you can show.

\n

- GPU depth. You have written or tuned kernels (CUDA, Triton or Warp).

\n

- Serving. Models in production on GPUs, with latency and cost per inference you measured.

\n

- Topology optimization and computational design. Generative and physics-driven design methods; watertight repair, remeshing and decimation.

\n

- Writing. A go/no-go memo a founder can act on.

\n

\n

NICE TO HAVE

\n

\n

- Computational mechanics: finite elements, contact, material models; GPU solvers (Warp, multigrid, Krylov methods).

\n

- Rust; CUDA Rust (cuda-oxide, cutile-rs).

\n

- Dataset curation and licensing for research models and weights.

\n

- Differentiable rendering, Gaussian splatting, multi-view diffusion.

\n

- B-rep and parametric CAD; CAD code generation.

\n

- Publications at CVPR, ICCV, SIGGRAPH or NeurIPS.

\n

\n

HOW WE WORK

\n

\n

Everyone runs several AI coding agents using our custom dev tooling within a proven process. You have the freedom to expand your own tooling and influence how we build and ship fast. Research depth and serving experience matter most for this role; prior experience with coding agents is not required.

\n

\n

We are fast, adaptable, and introspective of our work. We celebrate your wins, and feed your creativity. Ultimately, we all love what we are building and wouldn't waste our time doing anything else.

\n

\n

Expect our founders up at 2am sending memes while in flowstate building core product and making wild breakthroughs.

\n

\n

Remote across the US, with the company's home at Newlab in Detroit and occasional travel there.

\n

\n

PAY, EQUITY AND BENEFITS

\n

\n

- Equity, four-year vesting with a one-year cliff.

\n

- Health, dental and vision through our PEO.

\n

- US work authorization required. The role works with customer and partner data held under NDA and may require US-person status for export-control compliance.