Artificial intelligence (AI) relies heavily on data; however, a significant amount of health data remains untapped. This is largely due to concerns surrounding patient privacy, regulatory requirements, and intellectual property protection.
German entrepreneur Robin Röhm identifies this as “the fundamental challenge” in developing AI solutions within life sciences and pharmaceuticals. Additionally, he notes that collaboration involving sensitive data presents its own set of difficulties. Röhm’s startup, Apheris, seeks to overcome these hurdles through federated computing, enabling secure access to data for AI model training without relocating it, thanks to a decentralised methodology.
Among Apheris’s clientele are industry leaders such as Roche and various hospitals.
The fundamental principle of federated computing is that “computations occur locally where the data is stored, while only the results (for example, model parameters) are sent to a central system,” explains Marcin Hejka, co-founder and managing partner at OTB Ventures. Hejka has played a pivotal role in securing an $8.25 million Series A funding round for Apheris, in partnership with fellow deep-tech investor eCAPITAL.
Hejka envisions Apheris as a vital element in the emerging federated data networks. “We are witnessing the emergence of a mature ecosystem featuring third-party software tools such as open-source federation engines, data quality instruments, and security solutions,” he shared with StartupSuperb. “Apheris also facilitates smooth integration with other privacy-enhancing technologies, including homomorphic encryption, differential privacy, and synthetic data.”
The fresh funding follows a strategic shift for Apheris. Initially, Röhm and co-founder Michael Höh launched the company in 2019 with aspirations to develop a federated learning framework that competed with existing open-source solutions, influenced by their previous experience at Janus Genomics. However, following a significant seed round in 2022, they transitioned in 2023 to focus more on the interests of data owners, particularly in the pharmaceutical and life sciences sectors.
According to Röhm, this shift has proven fruitful. Apheris established a product-market alignment with its newly launched product in the last quarter of 2023, achieving a fourfold increase in revenue since. Backed by returning investors such as Octopus Ventures and Heal Capital, this latest funding round has lifted the company’s total financing to $20.8 million. These resources will be allocated towards recruiting senior professionals well-versed in life sciences, including on the commercial front.
The Apheris Compute Gateway, a software agent facilitating connections between local data and AI models, is currently operational with the AI Structural Biology (AISB) Consortium. This collaborative effort includes members like AbbVie, Boehringer Ingelheim, Johnson & Johnson, and Sanofi, all of which participate in AI-driven drug discovery efforts.
Apheris plans to concentrate on protein complex prediction as one of its key areas with the new investment. Although applicable across various use cases, it recognises its potential value in scenarios where public data is sparse, yet richer and more varied data exists that will only become accessible if life sciences companies feel assured about sharing it.
Röhm emphasises, “Unless we address the concerns of data owners regarding data utilisation for AI, we believe that the true potential of AI cannot be unleashed. This is at the heart of our mission.”
Leave a Reply