Flower Labs, a startup that offers open-source software enabling AI models to be trained without centralizing data, has reached a valuation of \(100 million after securing \)20 million in funding led by Felicis Ventures from Silicon Valley.
The technology developed by Flower facilitates the implementation of “federated learning,” a method allowing AI model training without consolidating all data on a central server. This approach enhances data privacy and security, making it particularly appealing to sectors like healthcare, finance, and defense.
The founders, Daniel Beutel, Taner Topal, and Nicholas Lane, initially collaborated at the University of Cambridge. Lane, a machine learning professor and former Samsung AI lab director, joined forces with Beutel, a PhD candidate, and Topal, a visiting researcher, to create the Flower framework for federated learning in 2020. As the framework gained traction among enterprises and government entities, they established Flower Labs in March 2023 to support the platform and develop tools for easier deployment in federated learning projects.
With over 1,100 projects utilizing the Flower framework, the startup has fostered a community of 3,100 developers on a dedicated Slack channel. Notable users include Accenture, Orange, and Siemens Healthcare.
The funding will be utilized to expand awareness of federated learning, increase developer adoption, and bolster the team, which currently consists of 13 members across Europe.
Federated learning operates by distributing the model to individual devices housing data, enhancing privacy as data remains localized. Despite potential vulnerabilities, additional techniques like differential privacy can further safeguard the training data.
While federated learning is advantageous for privacy and security, it is more suitable for smaller models due to limitations in deploying large models to edge devices. Nevertheless, the approach has gained traction due to regulatory compliance and the potential to leverage distributed computing power for AI training.
Flower Labs aims to democratize AI training by tapping into distributed computing resources, potentially revolutionizing the landscape where AI computing shifts from centralized cloud servers to edge devices like smartphones and laptops. This could address the current challenges of GPU scarcity and high costs associated with AI training.
