The core message of this session is that Germany can turn deep-tech research into globally competitive companies when research organizations deliberately connect long-horizon science with fast-moving spin-offs. The discussion opens by framing “physical AI” as AI gaining “bodies, arms and legs,” with robotics as the practical arena where this shift becomes visible. The moderator introduces Sven Parusel, whose robot arm work led to Franka Robotics, now part of Agile Robots, and Professor Karsten Lemmer from the German Aerospace Center, positioning the conversation between industrial robotics scale-up experience and institutional research depth. Agile Robots is described as a company based in Sandling with thousands of employees worldwide and a humanoid robot visitors can see at Bits & Pretzels at the German Aerospace Center booth.

Karsten Lemmer then explains DLR’s role as a research organization working across aviation, space, energy, transport, and defense, with 12,000 employees whose discoveries are converted into practice in several ways, including startups. He uses space exploration as an example of research opening “new perspectives on the future,” then connects that broader mission to spin-offs that turn such perspectives into reality. His central argument is that deep-tech startups are innovation drivers for Germany, but they are difficult because deep tech is time-consuming and costly. He points to more than 100 DLR startups since the 1990s, around 90% of which are still active in the market, as evidence that successful deep-tech commercialization is possible despite the challenge.

Sven Parusel describes his own path from DLR’s Institute of Robotics and Mechatronics, where he researched safe physical human-robot interaction, to Agile Robots, founded in 2018 to bring novel robotics technology into industry. He explains that the company began with the Diana 7 lightweight, sensitive robot arm, then expanded into hands, mobile platforms, robotic systems, and a humanoid robot. For Agile Robots, physical AI means more than making a single robot smarter: it means connecting robotic components through the agile core software stack so production lines can perceive their environment, adapt, learn from what they see, and become more intelligent factory-floor systems.

The session also emphasizes the strategic value of early research investment. Lemmer says DLR uses AI-supported foresight to analyze public data and cross-reference it with DLR research in order to identify strategic research areas ahead of the curve. He notes that DLR has researched robotics for more than 40 years and unveiled its first humanoid robot, Justin, in 2006. This long-term funding base enables spin-offs and partners to build on established expertise, while contributing to German and European sovereignty in space access, energy independence, automation, dangerous-task execution, and skilled-worker shortages. The provided transcript ends as Lemmer asks what Agile Robots’ humanoid Agile One offers at the Bits & Pretzels booth.