MELLODDY was a European-funded initiative designed to solve a long-standing challenge: how can pharmaceutical companies improve machine learning (ML) models for drug discovery without exposing sensitive data?
The project brought together 10 leading pharma companies — including Bayer, GSK, and Novartis — alongside key technical partners like NVIDIA, BME-HIT, Owkin, KU Leuven, the Substra Foundation, and Kubermatic. Together, they developed the first industry-scale federated learning platform for drug discovery, enabling companies to train ML models collaboratively while keeping proprietary data confidential.
Using federated learning, MELLODDY aggregated training updates — rather than raw data — from each partner after every iteration. This approach ensured data privacy while simultaneously developing more accurate predictive models. In fact, the global predictive model that resulted from this project outperformed all of the single partner models.
Kubermatic played a key role by providing the scalable Kubernetes infrastructure that allowed each pharma partner to run the platform securely on Amazon Web Services (AWS).
MELLODDY has set a new industry standard for AI-driven drug discovery. With the developments from this project, pharma companies will now be able to cooperate in training ML models, without compromising confidentiality — enabling a faster drug discovery process.
MELLODDY has set a new industry standard for AI-driven drug discovery. With the developments from this project, pharma companies will now be able to cooperate in training ML models, without compromising confidentiality — enabling a faster drug discovery process.
