Bifrost

Bifrost builds virtual worlds and synthetic datasets that artificial intelligence teams can use to train AI models for applications in areas such as gaming, the metaverse, mobility, robotics, space, and defense.

AI teams all over the world struggle to find high-quality labeled training data at scale. This is a major hindrance in developing functional machine learning systems. The challenge is particularly complex in highly visual applications such as defense, space, robotics, mobility, gaming, AR/VR, and CGI, requiring millions of instances and scenarios in their datasets. The current process of collecting such datasets from the physical world requires substantial infrastructure, large teams, millions of dollars, and takes years to complete.

With Bifrost’s tools, teams can now craft diverse, customizable, and rare datasets using synthetic, computer-generated data. This is a significant advancement from the standard workflow of physically collecting and processing data from the real world. Bifrost enables model training with higher accuracy within minutes, instead of the months or years it can usually take.

Bifrost’s data has been tested in mission-critical scenarios by sophisticated organizations like the NASA Jet Propulsion Laboratory and the US Southern Command. They deployed the datasets for challenging projects like validating Martian landing systems and detecting narcotic-carrying semi-submersibles in the Atlantic Ocean, respectively. Bifrost’s capabilities have also been used by leading players for aircraft detection, maritime domain awareness, vehicle detection, and environment generation, among other applications.

Bifrost was founded in 2020 by freshman-year friends, Aravind Kandiah and Charles Wong. Aravind has worked on AI research and computer vision for the past five years. He was the first engineer at Medios, a healthcare startup that created a low-cost AI-driven solution for diagnosing diabetic retinopathy that outperformed Google in clinical trials in India. Charles built electric race cars in university, was an autonomous vehicle engineer at Motional, and spent time as an analyst at Boost VC, where he built investment theses around quantum computing.