Eka Robotics

Eka Robotics is a robotics company developing force-aware systems for dexterous manipulation. Its Vision-Force-Action (VFA) model combines images with feedback from robotic hands during physical contact.^[2] The company was co-founded by Pulkit Agrawal and Tuomas Haarnoja and spun out of Agrawal's laboratory at MIT CSAIL.^[3]

Company and research background

Agrawal is an associate professor in MIT's Department of Electrical Engineering and Computer Science. His research group, the Improbable AI Lab, studies how robots learn perception, control, manipulation, and locomotion through experience.^[4][6] Haarnoja previously worked at Google DeepMind.^[3]

Haarnoja co-authored the 2018 Soft Actor-Critic paper with Aurick Zhou, Pieter Abbeel, and Sergey Levine. That work introduced an off-policy reinforcement-learning algorithm that optimizes expected reward together with policy entropy, and evaluated it on continuous-control benchmarks.^[5] It is part of the founder's research background, rather than a published specification of Eka's VFA implementation.

MIT described Agrawal's work on robotic hands and wrists as a move toward control based on forces rather than position alone. Its 2025 account also identified him as an Eka co-founder, before the company's October 2026 financing announcement.^[6]

Vision-force-action model

Eka designs its model, sensing, controller, and hardware together. The approach uses contact as information: slipping can prompt a grip adjustment, while resistance during insertion can prompt a change in motion. Vision provides information about the scene; force feedback provides information about interactions after contact begins.^[2]

The company describes generalization across objects, tasks, and environments as a goal for its VFA model.^[7] These are manipulation capabilities, rather than a claim that the system can carry out arbitrary physical work.

Training and transfer to hardware

Eka trains manipulation policies with reinforcement learning in physics simulation. The company says it varies objects, environments, and physical conditions to generate robot experience before transferring learned behavior to hardware.^[1] This makes sim-to-real transfer, particularly contact dynamics, a central part of its approach.^[2]

Demonstrations and performance claims

The October 2026 announcements describe several manipulation tasks, with different levels of evidence:

The throughput is a company-reported laboratory result, not an independently established production rate. Andreessen Horowitz identified maintaining performance throughout full production shifts as the next test. The picking figure should not be read as the rate for bulb installation or wire insertion.^[2]

Funding

On October 20, 2026, Eka announced a $150 million Series A led by Andreessen Horowitz and F-Prime. Obvious Ventures and E14 Fund, identified as its seed lead investors, continued their support.^[1]

The company also named MFV Partners, Bison Ventures, Glasswing Ventures, and Samsung Next as participants, alongside unnamed Fortune 500 strategic investors. Eka said the funding would support expansion of VFA to more tasks and accelerate customer deployments, with sustained reliability in real operations as its focus.^[1]

References

1. ^1 ^2 ^3 ^4 ^5Eka Robotics. "Eka Robotics Raises $150M to Build Superhuman Intelligence for the Physical World." October 20, 2026. Company announcement.
2. ^1 ^2 ^3 ^4 ^5Raghu Raghuram and Yoko Li. "Investing in Eka Robotics." Andreessen Horowitz, October 20, 2026. Investor announcement.
3. ^1 ^2Audrey Woods. "Startups at CSAIL: A Roundup of Recent Ventures." MIT CSAIL Alliances, August 31, 2026. University account.
4. ^Pulkit Agrawal. "Pulkit Agrawal." MIT CSAIL faculty research page. Accessed October 21, 2026. Faculty profile.
5. ^Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor." Proceedings of the 35th International Conference on Machine Learning, PMLR 80:1861-1870, 2018. Research paper.
6. ^1 ^2Alex Shipps. "Pulkit Agrawal receives the IROS Toshio Fukuda Young Professional Award." MIT CSAIL, October 24, 2025. University research background.
7. ^Eka Robotics. "Eka Robotics." Official website. Accessed October 21, 2026. Company overview.