Cortex AI nets $6m to collect real-world data for robots

Cortex AI, the robotics venture of Carousell co-founder Lucas Ngoo, has raised US$6 million in seed funding.

The round saw participation from companies like 500 Global, Golden Gate Ventures, Matrix Partners, Liquid 2 Ventures, and Y Combinator, as well as angel investors like Marathon Management Partners’ Gokul Rajaram and Opendoor CEO Kaz Nejatian.

Cortex AI founder Lucas Ngoo / Photo credit: CareSense

Ngoo, who stepped down from his managing role at Carousell in 2024, tells Tech in Asia that Cortex AI will use the new capital to build its team in Singapore and the Philippines. The firm is hiring for mechanical engineers, software engineers, robot operators, and machine learning engineers.

Ngoo will be based in San Francisco, where many of the company’s core clients, mostly AI frontier labs, are located. Without disclosing the names of its current customers, Ngoo says the company has US$100,000 in paid contracts so far.

“The data we want to collect is global,” Ngoo says, “But we leverage Southeast Asia for operations because of the familiarity and connections to small and medium businesses we built during the Carousell years.”

Brains over brawn

Instead of developing robots, Cortex AI focuses on the data that powers them.

“We don’t have the luxury of a ‘Wikipedia-scale’ dataset for robotics” unlike when training for large language models, Ngoo points out.

To help solve this problem, Cortex AI studies workers in industrial workplaces – such as factories, warehouses, and logistics hubs – and collects information on their hand movements.

While synthetic and robotics simulations from companies like Nvidia are available, Ngoo says Cortex AI wants to collect data that’s “more true to life.”

Image credit: Timmy Loen

“The real world is messy,” he says, “So the data has to be collected from the real world itself.”

Gathering this information are Cortex AI’s human operators, which the founder says will form the bulk of its teams in Singapore and the Philippines. The company then uses the data to train its model.