About HILOS

HILOS is building the first 3D design platform made specifically for footwear—no CAD required. Powered by machine learning, our core geometry engine breaks down barriers between 2D and 3D, design and manufacturing.

Our cross-functional team sits at the intersection of footwear and software, combining craft and intelligence to give designers a second brain and third hand. We’re here to amplify human impact in real-world creation—and we believe the tools should be just as inspiring as the work they enable.

The Role

We're looking for an AI Research Fellow who is deeply engaged with the fast-moving landscape of 3D generative models and eager to bring that knowledge into a production context. This is a research-heavy, hands-on role: you'll spend your time tracking, reproducing, and benchmarking the latest open-source 3D generation and reconstruction models — then working with our engineering team to integrate the most promising approaches into our platform.

The ideal candidate is someone who reads papers on Hunyuan3D, TRELLIS, Tripo, InstantMesh, and similar systems for fun — and has the engineering chops to get them running, evaluate them rigorously, and adapt them to real constraints like inference speed, surface quality, and downstream CAD compatibility.

What You'll Do

• Survey and benchmark SOTA 3D generation models — Track, reproduce, and systematically evaluate emerging open-source models (Hunyuan3D, TRELLIS, Tripo, InstantMesh, MV-Adapter, Stable Zero123, etc.) against our quality, speed, and manufacturability criteria.
• Build rapid evaluation pipelines — Develop reproducible benchmarking infrastructure for comparing 3D generation approaches on metrics like surface fidelity, mesh topology, inference latency, and texture quality.
• Prototype integration paths — Take the most promising models from benchmarking to proof-of-concept integrations with our platform, working closely with our AI/ML and computational geometry teams.
• Explore image-to-3D and multi-view pipelines — Investigate and improve pipelines that go from 2D concepts (sketches, photos, multi-view renders) to high-quality 3D assets, including multi-view diffusion, feed-forward reconstruction, and SDS-based optimization.
• Research reverse engineering approaches — Explore methods for extracting parametric or code-based design primitives from generated meshes or point clouds, bridging neural outputs and CAD-compatible representations.
• Contribute to internal research documentation — Write clear technical summaries, comparison reports, and recommendations that help the team make informed model and architecture decisions.
• Collaborate cross-functionally — Work with computational geometry, product, and design teams to ensure research outputs align with real user workflows and manufacturing constraints.

What We're Looking For

Must-haves:

• Active engagement with the 3D generative AI research landscape — you follow new releases, read papers, and have opinions on the tradeoffs between current approaches.
• Hands-on experience running and modifying open-source 3D generation or reconstruction models (any of: Hunyuan3D, TRELLIS, Tripo, InstantMesh, Zero123++, SV3D, CRM, Wonder3D, or similar).
• Strong Python and PyTorch skills, comfortable working with GPU infrastructure and debugging CUDA/model issues.
• Experience with 3D data representations — meshes, point clouds, SDFs, NeRFs, or Gaussian Splatting.