The spin-off is based on research in the field of explainable AI for high-frequency CNC machining data, which led to the development of the MVP (minimum viable product) anoMLix. This technology enables real-time quality inspection during machining without the need for additional sensors or reference workpieces.
Research has shown that defects, tool wear, and material anomalies can be detected directly from torque and force signals using a combination of signal processing, statistical modeling, and Tiny ML methods. A novel workflow has been developed to continuously adapt to tool wear and distinguish real defects from “anomalies in normal behavior” such as weak areas in a workpiece.
Compared to existing solutions, anoMLix offers an integrated approach that combines defect detection, material quality analysis, and tool life in an explainable platform. The algorithms were designed to automatically calibrate thresholds, ensuring robustness throughout the entire tool life cycle while maintaining interpretability for end users.

The research also introduced methods for predicting the remaining service life of tools, linking anomaly detection with wear progression models. Pilot tests confirmed that early fault detection can reduce unnecessary machining operations, thereby lowering energy consumption and CO₂ emissions. Initial benchmarks for anoMlix show CO₂ savings of up to 15–21% and cost savings of 26–32% in manufacturing.
These results form the basis for the spin-off and confirm both the technical feasibility and the economic and ecological potential of anoMLix for industrial manufacturing.