Antim Labs provides a simulation infrastructure platform for physical AI, delivering structured 3D environments, datasets, and evaluation suites that enable robotics teams to train and test embodied agents more efficiently. Their Gizmo workflow converts inputs such as point clouds, floorplans, prompts, or images into fully generated simulation worlds in formats like USD, MJCF, and SDF, supporting rapid scenario variation and reproducible failure analysis. This reduces the weeks‑long effort typically required to author custom robot simulations.
Funding
Funding not disclosed
Founders
Product
Problem
Creating simulation environments for embodied AI and robotics remains a manual, time‑consuming process, often taking weeks per scene, offering limited scenario variation, and producing fragile exports that are hard to reproduce.
Solution
Antim Labs offers a structured simulation infrastructure that automates the generation of 3D worlds from inputs such as point clouds, floorplans, prompts, or images. Their Gizmo workflow converts these inputs into simulation‑ready scenes in standard formats (USD, MJCF, SDF) with built‑in structure, articulation, and physical meaning. The platform also provides ready‑made datasets and evaluation suites that deliver pass/fail metrics, enabling teams to assess agent performance and reliably reproduce failure cases. By streamlining authoring, Antim Labs reduces environment creation time from weeks to minutes, allowing robotics teams to iterate faster and focus on model development.
Target Audience
Primary customers are robotics research labs, AI teams building embodied agents, and companies developing simulation‑based training pipelines for physical AI.
Features
- Automated world generation from diverse inputs (point cloud, floorplan, text prompt, image)
- Export of simulation scenes in industry‑standard formats (USD, MJCF, SDF) with articulated objects and physical properties
- Integrated evaluation suite delivering quantitative pass/fail metrics for embodied agents
- Ready‑to‑use datasets and benchmark scenarios tailored for robotics and embodied AI research
- Structured simulation environments that support reproducible failure analysis and iterative testing