The challenge of acquiring quality real-world data remains one of the largest obstacles for companies developing AI models, especially in applications that extend into the physical world. Obtaining clean, well-labeled data is a costly and labor-intensive process, creating a significant barrier for industries that rely on AI for robotics, automation, and other industrial operations.
Bifrost, a San Francisco-based startup, offers a solution through its 3D data-generation platform. Bifrost's technology enables companies to create simulated 3D environments that can accelerate the training of AI models, allowing robotic systems to quickly adapt to new objects, tasks, and environments. This simulation approach reduces model training time from months to mere hours. Recently, Bifrost raised $8 million in Series A funding, led by Carbide Ventures, to continue its work in this field.
According to Bifrost’s co-founder and CEO, Charles Wong, gathering the vast amount of data typically required for AI model training usually entails deploying large fleets of robots, collecting extensive hours of footage, and performing extensive data labeling and quality checks. Such a process can be prohibitively expensive and time-consuming. Bifrost’s platform offers a more efficient alternative, creating high-quality simulated data to enhance model training for AI in industries where real-world data collection is both costly and complex.
Founded in 2020 by Charles Wong and Aravind Kandiah, Bifrost was established to tackle the AI data challenges faced by robotics and industrial companies. Wong, with experience at the autonomous driving company NuTonomy, and Kandiah, who developed medical AI for detecting diabetic retinopathy, joined forces to address the essential need for high-quality data to advance AI systems in practical applications.
Bifrost’s approach differs from many current 3D simulation tools on the market. Unlike platforms such as Nvidia’s Omniverse, which require a dedicated 3D simulation team, Bifrost’s solution allows AI engineers to generate complex data without specialized 3D skills. This functionality is especially valuable for companies in heavy industries like defense and maritime that need adaptable AI systems capable of operating in unpredictable environments.
Currently, Bifrost’s platform is available in closed beta with a select group of industrial partners. With the new funding, the company plans to expand its platform, scale its staff, and prepare for a public launch. Although Bifrost’s primary market is the U.S., it is also gaining traction in Japan, a country with a significant industrial base. Revenue is generated through an annual subscription model, aimed at large industrial companies, government bodies, and advanced startups developing robotics and automation solutions.
Bifrost’s Series A funding, which brings its total raised capital to $13.7 million, saw additional investments from Airbus Ventures, Peak XV’s Surge, Wavemaker Partners, MD One, and Techstars. With 22 employees based in the U.S. and Singapore, the company’s near-term goals include expanding its applications for critical industrial use cases, with plans to broaden its platform to commercial robotics by 2025 as demand for robotic applications grows across major sectors.
