Eka Robotics
Category: Robotics / Embodied AI
Eka Robotics builds a Vision-Force-Action (VFA) foundation model and force-sensing gripper that let robots perform fast, general-purpose dexterous manipulation. Eka Robotics was founded in 2024. The company is led by Pulkit Agrawal. Based in Cambridge, Massachusetts, United States. Team size: 11-50. Total funding raised: $161M. Latest round: Series A. Key investors include Andreessen Horowitz (a16z), F-Prime, Obvious Ventures, E14 Fund, MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next.
- Founded
- 2024
- Headquarters
- Cambridge, Massachusetts, United States
- Team size
- 11-50
- Total funding
- $161M
Value proposition
Closes the generality-performance gap in robotics: one model that generalizes across objects, tasks and environments while running at production speed (about 1,500 picks/hour, roughly 2.5x a human) with over 95% reliability and safe human interaction.
Products and solutions
Vision-Force-Action (VFA) foundation model for manipulation, custom force-sensing robotic gripper/hand (reportedly under $1,000), simulation and reinforcement-learning training stack for sim-to-real transfer.
Unique value
Trains robots to master physics through large-scale reinforcement learning in simulation, combining vision and force sensing, rather than imitating human demonstration data. The result is superhuman speed with generalization to objects outside the training set.
Target customer
Industrial operators and manufacturers with pick-and-place, assembly and wire-manipulation work, plus the integrators that deploy robots on their lines.
Industries served
Manufacturing/assembly (incl. GPU server rack assembly), logistics/warehouse picking, food handling (e.g., delicate produce), industrial automation
Technology advantage
Force is a first-class modality, not vision alone. Physics-simulation RL generates thousands of hours of robot experience in minutes, avoiding costly teleoperation and human-demo data collection. A proprietary flexible gripper with dense force sensors handles contact-rich tasks (bulb screwing, wire insertion, gripping raspberries). The founders are leading RL-for-robotics researchers.
How they differentiate
Competitors mostly train vision-language-action models on human demonstration or teleoperation data. Eka instead uses a vision-force-action model trained by sim-to-real RL. Eka targets the low-level physical manipulation layer from the bottom up, where competitors focus on high-level planning, and prioritizes production-grade speed and reliability over breadth alone.
Main competitors
Physical Intelligence, Skild AI, Generalist AI
Key partnerships
Samsung Next and several unnamed Fortune 500 strategic investors, secretive industrial pilots for pick-and-place and wire manipulation
Major milestones
Founded 2024 per The Robot Report (founder LinkedIn shows Jan 2025 start). May 2025: $11M seed led by Obvious Ventures and E14 Fund. WIRED coverage of an early model, calling it a 'ChatGPT moment for the physical world'. Oct 2026: $150M Series A co-led by a16z and F-Prime at a $1.45B valuation, reaching unicorn status. Presented at Actuate 2026.
Growth metrics
$1.45B post-money valuation (Oct 2026); about 25 employees as of April 2026; robot picks about 1,500 objects/hour (up to 1 kg) at over 95% reliability.
Market positioning
Physical-AI unicorn focused on dexterous manipulation as the foundation-model layer for robot hands. Valued at $1.45B post-money, below larger generalist robot-brain players such as Physical Intelligence (~$11B) and Skild AI (~$14B).
Geographic focus
United States (Boston/Cambridge robotics hub); global industrial automation market
About Pulkit Agrawal
Associate Professor of EECS at MIT, leads MIT CSAIL Improbable AI Lab; Co-founder of SafelyYou (computer vision for dementia fall prevention); PhD UC Berkeley. Researcher in reinforcement learning, force-aware control, dexterous manipulation and sim-to-real transfer.
Latest news about Eka Robotics
More Robotics / Embodied AI companies
Official website: https://ekarobotics.com