Why We Invested in Trener Robotics: Building the Physical AI Layer for Industrial Automation

Industrial robots are widely deployed across manufacturing, but most are still programmed to perform a single, narrowly defined task. Despite millions of robots in operation globally, reconfiguring them for new parts, fixtures, or processes typically requires specialized engineers, custom programming, and significant downtime. As a result, automation remains slow, costly, and brittle, especially in high-mix production environments where frequent changeovers are the norm. Reprogramming a robot can take 60–200+ person-hours per changeover, and many factories still depend on system integrators for even minor adjustments. As labor shortages intensify and production variability increases, the gap between what robots could do and what they actually do continues to widen.

Trener Robotics is building a physical AI platform that enables industrial robots to learn tasks through training rather than hard-coded instructions. The platform combines vision, language, and action understanding to deliver pre-trained, production-ready robotic skills that adapt to real-world variability on the factory floor. Operators define tasks at a higher level, and the system translates that intent into executable robot behavior, validated in simulation and continuously improved performance over time with production data. 

Trener’s platform is built around reusable atomic skills that can be composed to perform a range of industrial tasks. CNC machine tending is the company’s initial deployment focus, chosen because it is a common workflow with high variability, frequent changeovers, and tight tolerances that expose the limits of traditional robot programming. By reducing setup time and enabling technicians, rather than robotics specialists, to reconfigure automation cells, Trener makes CNC automation more practical in real production environments. This use case serves as a wedge into a broader platform strategy, where additional skills can be trained and deployed over time, allowing the system to expand into other workflows as capability compounds.

From a macro perspective, manufacturing is undergoing a structural shift: labor shortages are persistent, product cycles are shortening, and factories are under pressure to produce more variants with fewer skilled workers. These pressures are exposing the limits of traditional industrial automation, which remains difficult to adapt when production conditions change. At the same time, advances in foundation models, simulation, and edge compute have made it possible to train robots in new ways, but most of these advances have yet to translate into systems that work reliably on the shop floor. Trener addresses this gap by delivering a platform that runs on existing robots, integrates with established factory systems, and can be deployed quickly with measurable returns. This focus on production readiness, combined with deep technical rigor, is what differentiates Trener from both brittle automation software and research-first robotics platforms.

One core differentiator for Trener is how the team is structured to balance long-term innovation with production deployment. The company has built a dedicated foundational R&D and AI research group, T-Labs, responsible for advancing the core embodied intelligence of the platform, including how robots perceive their environment, reason about tasks, and generalize skills across different setups and robot types. Alongside this, Trener maintains a separate team focused on customer deployments, system integration, and production reliability, ensuring the software performs consistently in real factory environments. From our perspective, this structure allows Trener to advance foundational capabilities without drifting away from the realities of industrial deployment, with live customer feedback directly shaping how the platform evolves.

This organizational design reflects the strengths of the founding team. Asad and Lars bring deep backgrounds in robotics and adaptive control, paired with a clear understanding of what it takes to deliver software that meets industrial reliability and OEM requirements. They have been deliberate in building both the technical foundation of the platform and the operational discipline required to deploy it through system integrators and robot manufacturers. Early engagement with leading partners across Europe and the U.S. has validated not only the technical approach, but also the go-to-market strategy.

At Geodesic Capital, we back founders building foundational technology that becomes core infrastructure for the next generation of industrial and enterprise systems. We are excited to partner with Asad, Lars, and the Trener Robotics team as they build and scale the platform.

Read Trener Robotics’ Series A announcement: Trener Robotics raises $32M Series A to bring Physical Intelligence to Industrial Automation, Providing a Foundational Intelligence Layer that Enables Software-Defined Control of Robots