Alex de Vigan, previously the founder of nfinite, has launched a new company called Physicl and secured $10 million in seed funding from Inovia Capital. The startup aims to repurpose the team’s prior 3D rendering and e‑commerce technology to create high‑fidelity simulations that help train robots and physical AI systems.
Why this matters for robotics
De Vigan’s former company, nfinite, focused on creating 3D product models for brands and retailers. As transformer‑based generative AI shifted that market, de Vigan saw an opportunity to apply the same rendering pipeline to robotics: producing simulated worlds with correct geometry, materials, and physics so that physical AI can better understand and manipulate space and objects. Researchers have found limitations in training robots from video data, and collecting real‑world labeled data via remote operations is costly; high‑fidelity simulations are a promising alternative if they accurately reflect real physics.
According to de Vigan, the pipeline his team developed can transform “unstructured, incomplete, heterogeneous real‑world information into a high‑fidelity physical representation,” and Physicl seeks to leverage that capability for robot training and world modeling.
Product: Studio
Physicl’s primary offering, Studio, ships with 10,000 3D environments and 100,000 assets representing a variety of objects. The product enables users to compose scenes and generate synthetic datasets aimed at specific training challenges. For instance, Studio can produce training examples to help a robot distinguish a cup on a cluttered countertop or when the cup is in shadow — the kinds of difficult edge cases that commonly break real‑world systems.
Early customer and workflow ambitions
World‑model builder SpAItial worked with Physicl as an early customer to improve simulation fidelity; SpAItial released its latest product, Echo‑3, last week. SpAItial’s tool creates 3D simulations from 2D imagery, and Matthias Niessner, SpAItial’s CEO and a professor of robotics at the Technical University of Munich, said the company needed high‑fidelity synthetic data like Physicl’s to train effectively.
Niessner described a workflow where synthetic data is generated on the fly according to the current training state — for example, a model or an orchestration system detects a deficiency and requests targeted synthetic examples, which are then autogenerated. Physicl positions itself not merely as a data supplier but as tooling to support ongoing, dynamic synthetic data generation at scale.
Market context and near‑term plans
Physicl enters a crowded field of startups and companies offering data collection, labeling, or synthetic generation tools for foundation model and robotics builders. De Vigan acknowledges the space is young and sometimes unclear about what will ultimately work. Physicl’s differentiator is its claim to convert large volumes of images or LIDAR captures into high‑quality 3D representations that can be used to generate billions of training permutations.
With the $10 million seed round, Physicl plans to expand its product capabilities and customer base while emphasizing the physical fidelity of its simulations as a competitive advantage.
Key facts
- Funding: $10 million seed from Inovia Capital
- Founder & CEO: Alex de Vigan
- Previous company: nfinite
- Product: Studio — 10,000 3D environments, 100,000 assets
- Early customer: SpAItial (product Echo‑3)
