Key Takeaways
- • Andreessen Horowitz (a16z) is leading the Series A financing in Eka Robotics, a company focused on dexterous robotic manipulation.
- • Eka has built a Vision-Force-Action model that combines visual perception with force feedback from its robotic hands, designing the model, sensing, controller and hardware together.
- • Eka trains manipulation policies through reinforcement learning in physics simulation, aiming to close the sim-to-real gap that has historically limited such approaches.
- • Eka reports up to 1,500 picks per hour in laboratory induction and bin-picking tasks, about 2.5 times faster than humans, though sustained production performance remains untested.
- • Co-founders Pulkit Agrawal (MIT CSAIL Improbable AI Lab) and Tuomas Haarnoja (co-creator of Soft Actor-Critic, formerly Google DeepMind) lead the team.
Summary
Andreessen Horowitz has announced it will lead the Series A round in Eka Robotics, according to its investment announcement. The firm frames dexterous manipulation, the ability of robots to perform hand-based tasks in variable real-world conditions, as one of the largest opportunities in robotics. The thesis centers on the observation that much of the physical economy, including manufacturing, electronics assembly, logistics and food production, still depends on human hands, and that industrial robots have historically worked only in highly structured environments where fixtures hold parts in place.
Eka's technical approach is a Vision-Force-Action model that combines visual perception with force feedback from its robotic hands. Rather than treating the model, sensing, controller and hardware as independent systems, the company designs them together so that physical contact itself becomes an information source. This allows a robot to correct a slipping grasp, locate objects through contact, or adjust motion during insertion, rather than halting on an error. The stated goal is to maintain performance across changing materials, objects and fits without engineers retuning the robot for each variation.
On the learning side, Eka trains manipulation policies through reinforcement learning in physics simulation, rather than relying primarily on human demonstrations. The company argues that simulation allows robots to practice millions of variations in object position, friction and contact, and to experience failure modes that humans rarely demonstrate. The announcement acknowledges that contact dynamics are difficult to model and that the sim-to-real gap has been one of robotics' hardest problems. Co-founders Pulkit Agrawal, a leading robot learning researcher at MIT CSAIL's Improbable AI Lab, and Tuomas Haarnoja, co-creator of Soft Actor-Critic at Berkeley and later a Google DeepMind researcher, are positioned as well suited to address that gap.
Eka reports up to 1,500 picks per hour in laboratory induction and bin-picking tasks, roughly two and a half times faster than humans, with demonstrated manipulation of fragile objects, slippery industrial parts and contact-rich insertion. The announcement notes that sustaining this performance through full production shifts is the next test. No financing amount was disclosed in the source material.
Eka's technical approach is a Vision-Force-Action model that combines visual perception with force feedback from its robotic hands. Rather than treating the model, sensing, controller and hardware as independent systems, the company designs them together so that physical contact itself becomes an information source. This allows a robot to correct a slipping grasp, locate objects through contact, or adjust motion during insertion, rather than halting on an error. The stated goal is to maintain performance across changing materials, objects and fits without engineers retuning the robot for each variation.
On the learning side, Eka trains manipulation policies through reinforcement learning in physics simulation, rather than relying primarily on human demonstrations. The company argues that simulation allows robots to practice millions of variations in object position, friction and contact, and to experience failure modes that humans rarely demonstrate. The announcement acknowledges that contact dynamics are difficult to model and that the sim-to-real gap has been one of robotics' hardest problems. Co-founders Pulkit Agrawal, a leading robot learning researcher at MIT CSAIL's Improbable AI Lab, and Tuomas Haarnoja, co-creator of Soft Actor-Critic at Berkeley and later a Google DeepMind researcher, are positioned as well suited to address that gap.
Eka reports up to 1,500 picks per hour in laboratory induction and bin-picking tasks, roughly two and a half times faster than humans, with demonstrated manipulation of fragile objects, slippery industrial parts and contact-rich insertion. The announcement notes that sustaining this performance through full production shifts is the next test. No financing amount was disclosed in the source material.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.