Machenta
Simulation-trained robot foundation models
CORE INFO
Machenta is developing robot-learning software for robotics companies, manufacturers, and robot operators. Its model is designed to let robots learn tasks such as packaging, assembly, machine tending, and tidying from a single physical demonstration.
- $40 million seed round co-led by Creandum and Union Square Ventures, announced October 27, 2026.businessinsider.com
- Company-run tests recorded 74/74 successful block placements, 78/78 screwdriver handovers, and 97/99 bottle-standing attempts; the tests were not independently assessed.robotharbour.com
- Business Insider reported that Machenta remains research-focused and has no near-term plans to commercialize its technology.businessinsider.com
WHY WE WOULD WORK AT MACHENTA
Redefine Robot Learning
Help make robots radically easier to teach: Machenta's vision-demo-action model aims to learn new physical tasks from a single demonstration, without task-specific retraining.
Build Frontier Physical AI
Work on simulation-trained robot foundation models, closed-loop vision-based action prediction, sim-to-real transfer, and cross-embodiment learning across robots, cameras, and grippers.
Research With Real Reach
Turn Imperial College research—including Instant Policy and work featured in Science Robotics—into systems that can handle packaging, assembly, machine tending, and manipulation.
Own a Foundational Role
Join an approximately 11-person research-and-engineering team where ownership, close collaboration, and in-person work shape the company's next chapter.
Join With Strong Momentum
Help scale a company backed by a $40 million seed round co-led by Creandum and Union Square Ventures, described as the largest seed round for a UK robotics startup.
Shape Practical Robotics
Work toward flexible automation for high-mix, changing tasks—from cable sorting and tidying to factory assembly—while solving the hard reliability and precision challenges of real-world deployment.
MARKET AND TRACTION
MARKET POSITION
COMPETITIVE ADVANTAGE
GROWTH TACTICS
✦ KEY METRICKEY METRICS
✦ KEY METRICNOTABLE CUSTOMERS
PRODUCT AND TECH
Vision-Demo-Action Model
Machenta's core robot foundation model observes a physical demonstration and ongoing camera input, then predicts actions for the robot. It is designed to teach new tasks without task-specific retraining.
Simulation-Based Pretraining
The model is pretrained entirely in simulated environments with large-scale randomized training. This approach aims to build general robot-learning capabilities while reducing dependence on real-world data collection.
Single-Demonstration Learning
Machenta's instant robot learning approach lets a person demonstrate a task once before the robot attempts it. Demonstrations can be provided through leader-follower control, VR controllers, kinaesthetic teaching, or direct hand guidance.
Cross-Embodiment Robotics
The pretrained model is designed to operate across different robots, cameras, grippers, and teaching methods. Demonstrated applications span manipulation tasks such as packaging, assembly, machine tending, tidying, and cable sorting.