Eka Robotics has announced a $150 million Series A round to accelerate development of its Vision-Force-Action model. The US startup says its AI allows robots to handle a wide range of objects at a rate of around 1,500 operations per hour.
Founded in 2024 and based in Cambridge, Massachusetts, Eka Robotics develops artificial intelligence for precision robotic manipulation. Its approach combines vision, force sensing and reinforcement learning to teach robots how to interact with objects and changing environments. The company plans to use the funding to develop its technology and expand deployments with customers.
Grasping objects, screwing in a light bulb, inserting wires: what Eka’s AI can do
Eka Robotics’ Vision-Force-Action (VFA) model is designed to let robots adapt their movements to the physical properties of the objects they handle. Rather than only replicating recorded human movements, the system learns from simulated physical interactions.
Eka says robots running its AI can grasp a wide variety of objects, including objects that were not in the training data. The company reports a rate of around 1,500 manipulation operations per hour for these grasping tasks.
Google DeepMind trains Apollo humanoid robot with Gemini 4 Argon: 7 successful throws from video
The startup also says its robots can screw in a light bulb and insert electrical wires. These capabilities are presented as results of its technology and do not constitute an independent measurement of its performance.
The aim is to combine two properties that matter for industrial applications: the ability to handle varied situations and fast, repeated execution. Eka is targeting dexterous manipulation tasks in particular, which require precise adjustments to movements and to the forces applied.
How Eka trains its model in simulation to learn object physics
What sets the VFA model apart is its training through simulation and reinforcement learning. Eka Robotics aims to give robots a form of physical intuition, based on analysing visual information together with the forces encountered during a manipulation.
According to the company, its simulation environment can generate thousands of hours of robotic experience in a few minutes by varying objects, environments and physical conditions.
NVIDIA launches Flora, three AI models that annotate robot videos to speed up training
This method exposes the model to a wide range of situations virtually before it is used on real machines. Eka says the results of its VFA model show fewer of the difficulties usually encountered when transferring what is learned in simulation to the physical world.
The company sums up its approach with two principles: large-scale reinforcement learning improves performance, while diversity of experience helps the model generalise to new tasks.
Who invested: Andreessen Horowitz and F-Prime lead the Series A
The $150 million Series A round is co-led by Andreessen Horowitz (a16z) and F-Prime, a venture capital fund affiliated with FMR LLC, the parent company of Fidelity Investments.
Obvious Ventures and E14 Fund, both investors in the seed round, are backing the company again. MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next and several strategic investors from Fortune 500 companies are also taking part.
Raghu Raghuram, general partner at a16z, says Eka’s Vision-Force-Action model brings together the versatility and speed that robotic deployments require. He also points to the team’s ability to combine robotics research with commercial development.
Shyam Kamadolli, managing director at Fine Structure Ventures, notes that a robot’s industrial value depends on its success rate and speed over thousands of repetitions. He sees learning physical properties as a way to move beyond the limits of systems trained mainly on human demonstrations.
What the funding will be used for: customer deployments
With this funding, Eka Robotics plans to accelerate development of its AI model and the rollout of its technology to customers. The announcement concerns an AI system for robotic manipulation, not the commercial launch of a specific humanoid robot model.
Eka’s industrial strategy relies on using this learning approach to carry out complex manipulations under variable conditions. The company aims to scale its grasping and dexterous manipulation capabilities to a level suited to repetitive operations at customer sites.