Twincher is a new class of trainable inverse solvers designed to help AI systems perceive and act in the physical world.
Many real-world tasks – perception, control and decision-making – require solving inverse problems: inferring hidden causes from observations, such as estimating an object’s pose from a 2D image or determining the actions needed to reach a desired state. Current approaches typically rely on iterative optimization or large learned models, which can be computationally expensive and difficult to make reliable across tasks.
Our approach goes beyond purely loss-driven learning. Guided by geometric principles, a twincher learns a representation space in which iterative search is steered toward the correct solution by construction rather than by chance. Matching the forward model then converges within a few iterations, opening a path toward single-step, noise-robust solutions to inverse problems. More broadly, we are developing deterministic, noise-resilient inference mechanisms for Physical AI systems.
Our first application prototype brings this approach to industrial quality inspection, recovering CAD parameters directly from LiDAR measurements.
From CAD to Quality –
Without Manual Tuning
LiDAR-Fit turns raw LiDAR scans into CAD-accurate deviation analysis that is robust to noise, misalignment and production variability.
Built on twincher inverse solvers, the system converges to the correct alignment without getting trapped in false minima, so residual errors stem from measurement noise and model mismatch rather than from optimization artifacts.