Bleu Robotics closed a €5M pre-seed led by Contrarian Ventures and OVNI Capital, with participation from Kvanted, WEPA Ventures, Motier Ventures, and Charlie Songhurst. The Paris-based startup will use the capital to grow its team and scale deployments across industrial partners.

The company's software uses abductive learning instead of end-to-end models. Operators teach robots new tasks at the workstation using handheld training grippers — no coding required. The system infers task intent and constraints from as few as ten demonstrations, maintaining the 99%+ accuracy industrial production lines require. Bleu claims this eliminates the need for vast training datasets and allows the same robot to be retrained as production needs shift.

The timing reflects a gap in the humanoid robotics stack. While hardware has advanced significantly, software deployment remains slow and data-intensive. Most competitors pursue general-purpose models trained on millions of data points — an approach Bleu argues doesn't suit manufacturing, where internet-scale robot data doesn't exist. Bleu's model is hardware-agnostic and already deployed with two undisclosed industrial partners.

The funding positions Bleu in a growing race to solve the "last mile" of humanoid deployment. If the one-hour deployment claim holds in real-world production environments, it would remove a major barrier to adoption in high-mix manufacturing — a segment where Europe and the US are attempting to reshore capacity. The key risk is whether ten-demonstration learning can generalize across edge cases and unstructured production anomalies at scale.