Bifrost helps industrialists speed up model training with its 3D data generation platform

For many companies working on AI models with applications in the physical world, data represents the biggest opportunity. This is also the biggest hurdle they face, as beautifully labeled and clean real-world data is available like hen’s teeth, and the cost and effort required to collect and clean the data can be enormous.

Bifrostplatform for generating 3D data, believes its technology can help robotics and industrial companies solve at least one part of this problem: the time it takes to train AI models. The San Francisco-based startup says its platform allows companies to create simulated 3D worlds to instruct their artificial intelligence models and help their robots adapt to new objects, tasks and environments in hours rather than months.

The company said Wednesday it has raised $8 million in a Series A funding round led by Carbide Ventures.

“Most of our customers need massive amounts of real-world data to train AI models,” co-founder and CEO Charles Wong said in an exclusive interview with TechCrunch. “This often means they will have to deploy fleets of robots in hundreds of locations, collect millions of hours of video footage, manually label data, and implement stringent quality checks to reduce human error and bias. This approach is cruel. It costs millions, takes years, and is virtually impossible to scale.”

Wong co-founded the firm with Aravind Kandia in 2020. Wong previously worked on artificial intelligence perception models for self-driving cars at NuTonomy, an MIT affiliate that worked on self-driving vehicles and autonomous mobile robots. Meanwhile, Kandia previously created an artificial intelligence medical system that detects early signs of blindness and diabetic retinopathy.

“It didn’t take us long to realize one fundamental thing: AI and robotics require huge amounts of high-quality data to function properly. And this data is very important,” Kandia told TechCrunch. “This can make or break the performance and potential of these systems. So we joined forces to launch Bifrost with a single goal: to solve the data problem so that AI and robotics can finally tackle the complex problems of the physical world.”

Bifrost says it differentiates itself from its competitors because creating such data for its platform does not require a team skilled in creating 3D modeling. This, according to Wong, gives AI engineers a significant advantage, allowing them to develop AI systems for tasks such as patrolling disputed waters with autonomous boats without having to hire a 3D team.

“In contrast, Nvidia Omniverse tools require a dedicated 3D team just to operate,” Candia said, adding that Bifrost allows engineers across a variety of heavy industries to teach AI systems new skills and achieve more faster.

The Bifrost product is currently in closed beta testing with select heavy industry partners. The startup will use the fresh money to fund a public launch of the platform in the coming months, as well as hire more employees to speed up product development.

Wong said the company’s main market is the United States, but it is also gaining traction in Japan thanks to the country’s sizeable industrial sector. The startup generates revenue through an annual subscription model.

The platform’s primary users are artificial intelligence developers who specialize in creating robotics systems, computer vision and perception models for applications in industries such as robotics, aerospace, defense, maritime, geospatial and industrial automation. Its target customers include large industrial companies, government organizations and late-stage startups, all of which will have teams focused on developing physical AI solutions in their respective fields.

“We are initially focusing on mission-critical applications in heavy industry. By 2025, we aim to expand the availability of the platform. … Looking to the future, we plan to support a wide range of commercial use cases for robotics, especially as robotics applications are rapidly emerging across virtually every major sector and industry,” Candia said.

The Series A brings Bifrost’s total capital to $13.7 million. Airbus Ventures, Peak XV’s Surge, Wavemaker Partners, MD One and Techstars also participated in the round. The company has 22 employees in the United States and Singapore.