The robot arm knows how to pick up a part. The humanoid knows how to walk. Neither of them knows how to run a laser engraving machine. Teaching them that, in a real factory, is a months-long project of custom programming, sensor calibration, and safety testing. It is the kind of problem that keeps expensive robots bolted to a single, simple task for their entire working lives.
Bleu Robotics, a Paris-based startup, thinks it can shrink that deployment timeline from months to under an hour. Its product is not a new robot, but a layer of AI software designed to teach existing robots,humanoids, mobile manipulators, industrial arms,new industrial tasks from just a handful of human demonstrations. The company recently raised a €5 million pre-seed round led by Contrarian Ventures and OVNI Capital to prove its first model, Bleu_0.1, can work outside the lab [Sifted, October 2026].
The deployment wedge
Bleu Robotics is betting on the deployment layer as its commercial wedge. The logic is straightforward: the hardware is getting better and cheaper, but the software to make it useful in dynamic, real-world settings remains a bottleneck. A factory might buy a versatile robot, but if reprogramming it for a new product line requires a team of specialist integrators for twelve weeks, the flexibility is theoretical.
The company’s approach is few-shot imitation learning. In a demonstration at VivaTech, its software guided a Unitree G1 humanoid through a complete laser engraving workflow on an xTool machine. The robot opened the machine’s door, positioned the part, initiated the engraving cycle, and retrieved the finished object [Roman Klokov - Bleu Robotics | LinkedIn]. The same software was shown operating a Franka robotic arm, suggesting an ambition to be hardware-agnostic [LinkedIn]. The promise is not that the robot understands the concept of engraving, but that it can reliably replicate the demonstrated physical sequence.
The research pedigree
The technical foundations come from co-founder Jean-Baptiste Mouret, a leading researcher in physical AI and robot learning, and Serena Ivaldi, a Director of Research at Inria who leads the Human-centered robotics team. The company says its work draws on two decades of robot-learning research from Inria, with which it maintains a partnership [bleu-robotics.com/index][Serena Ivaldi on LinkedIn]. This is not a product built from scratch in a garage; it is a commercial vehicle for academic research that has been maturing for years.
The other critical co-founder is Benoît Berkoukchi, who is presented as bringing the “reality of the factory floor” to the team [bleu-robotics.com/index]. The fourth listed co-founder is Jean-Baptiste Murphy. The blend is classic: deep technical credibility paired with an operator focused on the end customer. Their first major hire appears to be a Chief of Staff, and they are actively recruiting research engineers in robot learning and computer vision, signaling a build-up of technical muscle [Ashby].
Pre-seed | 5 | M EUR
The path to a Tier 1 factory
The stated target is Tier 1 manufacturing facilities, a segment where the pain of integration is high and the potential payoff from flexible automation is even higher [Ashby]. The company’s job postings explicitly mention deploying robots in these environments. The current challenge is moving from a compelling conference demo to a paid, production deployment in a noisy, unpredictable factory.
The risks here are not small. They are the fundamental challenges of robotics:
- The simulation-to-reality gap. A task learned in a controlled demo setting must work perfectly amid factory-floor vibration, variable lighting, and unexpected obstacles.
- The variability problem. Few-shot learning implies the robot can generalize from a small number of examples. If every slight variation in part placement or machine model requires a new demonstration, the efficiency gain evaporates.
- The safety imperative. In a research setting, a robot can drop a part. On a factory floor next to human workers, it cannot. Proving the system’s reliability and safety to industrial customers will be a steep hurdle.
The company’s answer, for now, is focus. It is not trying to solve general intelligence. It is targeting specific, repetitive industrial tasks where the sequence of motions is well-defined and the environment can be somewhat constrained. The laser engraving demo is a perfect example: a precise, multi-step procedure that is tedious for a human but torturous to program into a robot the old way.
What the checkbook says
The €5 million pre-seed is a substantial first round for a deep tech startup, especially one based in Europe. Lead investor Contrarian Ventures is a climate tech fund, which suggests it sees Bleu’s potential to drive efficiency and perhaps reduce waste in manufacturing. OVNI Capital is a French fund known for backing technical founders. The money likely buys 18 to 24 months of runway to turn the research prototype into a sellable product and land the first few pilot customers.
A back-of-the-envelope calculation illustrates the economic bet. If a traditional systems integrator charges €200,000 and takes three months to deploy a robot for a new task, the cost is that fee plus the opportunity cost of the idle robot and delayed production. If Bleu’s software can do it in an hour for a €20,000 annual license, the value proposition is stark, even if the software only works for a subset of tasks. The incumbent it must beat is not another AI startup, but the entrenched, manual process of robot programming and integration.
The next twelve months will be about validation beyond the trade show floor. The key milestone to watch is a signed pilot with a named industrial customer, where Bleu_0.1 is tasked with deploying a real, revenue-generating task. If they can show that their hour-long process holds up over a three-month production run, the bet starts to look less like research and more like a business.
Sources
- [Sifted, October 2026] The French robotics startup promising to deploy industrial humanoids after just 20 minutes of training | https://sifted.eu/articles/french-robotics-startup-industrial-humanoids-20-minutes-of-training
- [bleu-robotics.com] Bleu Robotics company website | https://bleu-robotics.com/index
- [Roman Klokov - Bleu Robotics | LinkedIn] Post on live demonstration at VivaTech | https://www.linkedin.com/in/rklokov/
- [LinkedIn] Post on launch of Bleu Robotics and Bleu_0.1 | https://www.linkedin.com/posts/beno%C3%AEt-berkoukchi-539a4b11b_today-im-launching-bleu-robotics-and-introducing-activity-7439591360598482944-14XK
- [Serena Ivaldi on LinkedIn] Profile information | https://www.linkedin.com/in/serena-ivaldi
- [Ashby] Bleu Robotics careers page | https://jobs.ashbyhq.com/bleu-robotics