Imperial College Spinout Machenta Lands $40M Seed Round

Your daily brief on leadership and business

Machenta, a London robotics startup founded by researchers from Imperial College London, has raised a $40 million seed round. Creandum and Union Square Ventures co-led it. The company told Business Insider about the round exclusively and declined to disclose its valuation.

It is building an AI model meant to teach robots new physical tasks after seeing them done just once.

Simulated data only

Rivals building general-purpose robot models are racing to collect real-world data. Machenta goes the other way. CEO and cofounder Edward Johns, an Imperial professor, said the company trains its model in computer-generated environments that mimic how objects move and interact. It uses, in his words, not a "single pixel of real-world data."

Once trained, the model can direct a robot through a new task without further training, Johns said. Examples are tidying a table or putting objects into a box. In a video shared with Business Insider, three robots pack and close a box on an assembly line. Johns said they learned the task this way.

The startup has 11 employees and is focused on research, with no near-term plans to commercialize the technology. The model can learn some tasks immediately. More precise tasks and complex movements remain challenging.

Crowded and well-funded field

Machenta faces rivals valued in the billions. Physical Intelligence is valued at roughly $11 billion. FieldAI is raising at a $10 billion valuation, Business Insider reported this month. Skild AI is valued at over $14 billion and already sells technology with similar capabilities.

In August, Skild unveiled a model it said could perform previously unseen tasks lasting up to 10 minutes after a single video demonstration. Last month it said it had reached $100 million in annual recurring revenue from more than 60 customers.

Johns began looking at this problem six years ago in his Imperial lab. Researchers physically guided a robot through a task, such as picking up a spatula and scooping an object. The robot then spent about 30 minutes collecting extra data about the object and its surroundings before it could do the task alone.

Johns and former Ph.D. student Vitalis Vosylius spent the next few years trying to remove that step. They founded Machenta late last year, after early signs that the model could perform new tasks right after a demonstration.

Johns doubts that more data is the answer. It also remains unclear how much training data general-purpose robots will need. Johns said Machenta is taking on the harder job of designing algorithms that generate useful synthetic data.

The funding will pay for a larger team and for training the model with more computing power and simulated data. The company hopes to teach robots more precise and complex tasks.

Source: Business Insider