Two founders, honest about what's real.
Early access. Two founders (TUM / LMU). Every number on this page is measured and reproducible — where we are weak, it says so.
A robot fault costs the same hour twice: once to notice something is wrong, and again to find out which of two hundred channels is telling the truth about it.
Self-tests only catch what someone thought to name in advance. The faults that stop a fleet are the ones with no name — a driver leaking DMA mappings, a clock stepping backwards, a gear ratio off by a factor nobody wrote down.
We split the work the way it actually divides: deterministic maths decides whether something is anomalous, and an agent that can read your channel names decides what it means. Neither half works alone.
- nowRuns today, measured on real recordings.
- measuredA number we reproduced — the method is written down.
- not yetNamed because it is missing, not because it is coming soon.
Two builders, honest about what's real.
AI systems and agentic workflows. Builds the detection engine and the agent layer that reads arbitrary channel names.
LinkedIn →
Aerospace and CAD at TUM. Builds ingestion, the live ROS 2 path and the robot-side deployment.
LinkedIn →