Eka Robotics is developing general-purpose robots for physical work. Its Vision-Force-Action model combines visual perception with force feedback, with an initial focus on dexterous manipulation.
In their October 2026 investment memo, a16z’s Raghu Raghuram and Yoko Li make the case for robots that can adapt to changing objects, materials and contact conditions. They point to Eka’s joint design of its model, sensing, controller and hardware as a way to turn physical contact into useful information.
The memo highlights reinforcement learning in physics simulation and the founders’ complementary robotics research. It frames production reliability as a test still ahead: sustained performance through full shifts remains to be demonstrated. a16z co-led Eka’s Series A alongside F-Prime.
WIRED’s Will Knight visited Eka’s office in Kendall Square, Cambridge, Massachusetts. His report describes a robot grasping a light bulb and screwing it into a socket.
Pulkit AgrawalCo-FounderMIT associate professor researching perception, control, robotics and reinforcement learning. Earned his PhD at UC Berkeley and previously co-founded SafelyYou.LinkedIn ↗
Tuomas HaarnojaCo-FounderCo-created Soft Actor-Critic at Berkeley and later worked at Google DeepMind on humanoid behaviors trained in simulation and deployed on physical robots.LinkedIn ↗
Kristian HartikainenFounding Engineer & Head of SoftwareCompleted his PhD at Oxford in reinforcement learning and robotics, with research internships at DeepMind and Google Brain Robotics and earlier software work at Statwing and Qualtrics.LinkedIn ↗The Machine Learning / Reinforcement Learning Engineer role includes deploying, evaluating and debugging policies on physical hardware with the robotics team. This Boston Area opening is listed on-site.
Eka’s launch announcement describes a team spanning robotics and machine-learning research, with experience at MIT, Berkeley, DeepMind, Boston Dynamics and Microsoft. The company emphasizes giving people trust, freedom and support to do their work.
The ML / RL engineering posting explicitly calls for clean code, version control and reproducible experimental workflows alongside developing learning algorithms.
After working with Pulkit on reinforcement learning and robot learning at MIT, Srinath Mahankali joined Eka. In May 2026, Pulkit described Srinath as the second teammate to join the mission and recognized their years of collaboration.
The Developer Productivity Engineer owns Bazel builds, CI and reproducible environments. The role works directly with researchers and robotics engineers to find recurring friction and measure whether tooling changes help.
Ryan Clough, Eka’s first recruiting hire, described joining a team offsite during his first week.
Pulkit invited robotics practitioners to Eka socials at ICRA in June and RSS in July for discussions about startup and research life. The ICRA invitation described rooftop drinks in Vienna, co-hosted with Ketryx, with no set agenda.
Eka’s Head of People opening asks a hands-on leader to build lightweight people systems and support first-time managers. The mandate includes improving onboarding and developing performance and feedback practices as the company grows.
The Forward Deployed Robotics Engineer troubleshoots systems at customer sites and demonstrations, then brings field findings back to engineering. The posting requires travel for 50% or more of the role.
