Bifrost raises $8M to accelerate AI's ability to solve problems in the physical world

2024-11-01 Bifrost AI, Inc HaiPress

SAN FRANCISCO,Nov. 1,2024 -- Bifrost AI,a Generative 3D Data platform,raised $8 million in Series A funding to meet the growing demand for AI training data and 3D simulations. Its physically accurate data generation engine provides the vast real-world scenarios crucial in rapidly training,testing and adapting models in AI and Robotics. The round was led by Carbide Ventures,with participation from Airbus Ventures,Peak XV's Surge (formerly Sequoia Capital India & SEA),Wavemaker Partners,MD One,and Techstars. The capital will be used to accelerate the commercialization of Bifrost's product for heavy industries such as aerospace,maritime,manufacturing,and national security.

Today's pressing challenges,such as labor shortages and escalating geopolitical tensions,primarily exist in the physical world. While AI has thrived in the digital realm,its potential in the physical domain remains untapped due to a lack of domain-specific trainingdatasets. More than 95% of enterprises struggle to obtain high-quality training data at scale,often pouring millions into sensor deployments and dedicating months to manual data labeling—only to face uncertain outcomes. This struggle is one of the biggest barriers to effectively implementing AI systems in the physical world.

Bifrost's software platform changes this by allowing teams to generate datasets in minutes using simulated 3D worlds. This allows teams to build new AI capabilities faster than ever,teach robots how to solve real-world problems,and adapt to new objects,tasks,and environments in hours instead of months.

Bifrost's breakthroughs in Generative AI and 3D graphics allow users to create realistic,industry-specific data & scenarios at scale—no 3D expertise needed. Designed for AI developers,the platform offers fully controllable real-time 3D worlds in supercharged Jupyter notebooks. Traditional 3D simulators often fail to deliver realism,while 2D generative AI lacks the nuanced control necessary for sophisticated applications. Bifrost bridges this gap with a hybrid approach,combining an unprecedented level of precision with realism. Users harness a powerful Python library to program every element of their 3D worlds,from camera movements to environmental conditions,automating and fine-tuning their simulations with ease.

This has enabled their customers to build new Physical AI capabilities faster than ever. Autonomous boats that can safeguard shipping lanes from piracy,industrial robots that can assemble new parts with unmatched efficiency,and drones that can inspect thousands of electrical grids for defects—all without human intervention. These systems serve as a vital link that allows digital systems to engage with the real world.

After LLMs and image generators,Physical AI represents AI's most important frontier yet. However,building these systems is incredibly hard. Bifrost is making this process faster,cheaper,and more reliable,so teams can focus on what matters: solving real-world problems with AI.

"We work with some of the most technically demanding organizations in the world,solving some of the toughest problems. This includes major U.S. government organizations and leading enterprises in heavy industry. One example is NASA JPL,with whom we have been collaborating on several initiatives,including data generation engines for Moon and Mars exploration," said Charles Wong,Co-founder and CEO of Bifrost AI.

"The market for robotics and autonomous systems is rapidly increasing. In the future,AI development will rely increasingly on synthetic data instead of real data. We believe the Bifrost team is uniquely positioned to enable this opportunity by bridging the data,allowing systems to be trained more efficiently and quickly by an order of magnitude. We expect their approach will unlock new use cases and opportunities,enabling AI to transform our physical world",said Stephen Lee,General Partner at Carbide Ventures.

"As AI becomes increasingly prevalent,actively deployed across commercial and defense-related applications at an unprecedented scale,Bifrost stands out as a trusted partner,ready to support teams working on advanced robotics and autonomous systems," adds Thomas d'Halluin,Airbus Ventures Managing Partner. "We are confident that Bifrost AI will unlock exceptional value at the intersection of 3D generative AI,advanced simulation,and design,which is already being used by major U.S. government organizations and leading global enterprises in heavy industry."

"One of the fundamental challenges in AI and robotics is generalization—the ability to learn new skills and handle a variety of scenarios and tasks. Traditionally,adapting to new tasks could take months. But with Bifrost,that timeline has shrunk to just hours. This acceleration enables teams to build AI systems that can tackle a wider range of problems,unlock larger markets,and generate more revenue with less upfront data acquisition cost." Aravind SK,co-founder and CTO of Bifrost AI.