Eka Robotics hits $1.45B valuation with force-sensitive AI

Founded by Pulkit Agrawal and Tuomas Haarnoja, Eka Robotics is moving away from the industry standard of vision-language-action models. Instead, the team has built a vision-force-action system. This approach allows their robots to handle fragile items—like a raspberry—with a level of dexterity that has stunned veteran roboticists. The system processes contact and pressure in real-time, enabling the hardware to adapt to novel objects without requiring constant retraining.

The performance metrics are aggressive: the Eka system can manipulate roughly 1,500 objects an hour, significantly outpacing human speed while maintaining a 95% success rate. By leveraging high-scale physics simulations rather than relying solely on human-recorded video data, the company is generating thousands of hours of training experience in mere minutes. This bottom-up approach to physical intelligence aims to bridge the gap between high-level AI planning and the messy reality of industrial tasks.

Despite the excitement, the transition from controlled laboratory environments to rugged industrial assembly lines remains the ultimate test. Experts like Kris Hauser note that while the current "pick and place" results are impressive, the technology must now prove it can operate reliably over millions of duty cycles in commercial settings. For now, Agrawal and Haarnoja are focusing on specific industrial verticals, betting that mastering the physics of touch will provide the foundation for a new generation of truly general-purpose machines.