Liquid AI, co-founded by robotics luminary Daniela Rus, is building general-purpose AI systems powered by liquid neural networks. Liquid neural networks use neurons governed by equations that predict each neuron's behavior over time, giving the architecture flexibility inspired by roundworm brains. Because of that design, these models are much smaller than traditional AI models and require far less computing power to run. Liquid AI plans to develop tailored liquid neural networks for applications in e-commerce, consumer electronics, and biotech. The company recently raised $250 million and is reportedly valued at over $2 billion. As part of the investment, Liquid AI will work with AMD to optimize its models for AMD GPUs, CPUs, and AI accelerators. Liquid AI develops models built on liquid neural network architecture, which the team says are much smaller, more compute-efficient, interpretable and able to adapt over time compared with typical large generative models. The company has demonstrated the architecture on tasks such as drone navigation, where liquid models generalized to unseen scenarios and outperformed other algorithms. Liquid AI plans to commercialize the architecture by building liquid foundation models beyond GPTs and by offering on-premises and private AI infrastructure plus a platform for customers to build domain-specific models. The team includes co-founder Daniela Rus and founding members Ramin Hasani (CEO), Mathias Lechner (CTO) and Alexander Amini (chief scientific officer). The company currently has a 12-person team and expects to grow to about 20 by early next year. Liquid AI targets applications ranging from drone search-and-rescue, wildlife monitoring and delivery to sequential-data domains such as power grids, medical readouts, financial transactions and severe weather prediction.

Total raised

$288M

Funding rounds

2

Latest round

Series A

Latest activity

Dec 2024