The Problem We Solve

Genomic and health data are siloed, sensitive, and hard to use collaboratively. Privacy risks, regulatory fragmentation, and technical incompatibility block researchers from pooling datasets — especially across borders.

Without secure infrastructure, the promise of AI in (genomics) research remains locked behind legal and technical walls.

The CoMPai Solution

CoMPai (Confidential Multi-Party AI) is a secure computing framework developed by OASYS NOW and validated with Erasmus MC, EPFL, NVIDIA, and Google. It adds a confidential aggregator to standard federated learning pipelines, preventing model weight leakage and ensuring full end-to-end trust.

Compatible with interpretable deep learning frameworks like GenNet, CoMPai enables AI-driven discovery — while keeping genomic data secure, auditable, and compliant.

How It Works

• 🧑‍🔬 Create or join a research project

• 📁 Prep your data (HDF5 supported)

• 🧠 Choose your model topology & prior knowledge embedding

• 🔐 Configure secure federation parameters

• 📊 Train across sites — without moving raw data

→ Supports both central and decentralized model setups

Privacy and Compliance

• Model updates are encrypted and never exposed to the coordinator

• Full support for confidential computing, Privacy Enhancing Technologies (PET)

• Future-proof for EHDS SPE frameworks

Research Validated

• ✅ Reproduced GenNet experiments (Nature 2021) in secure, federated form

• 🔄 Same results, no central data movement

• 🔍 Expert validation with Erasmus MC, NVIDIA, Google, EPFL

• 🧬 Follow-up projects with Alzheimer’s Disease Genomics Consortium (ADGC)

Use Cases

• Cross-institution AI training on genomics and health data

• Federated GWAS and genotype-phenotype prediction

• Secure benchmarking and reproducibility of ML models in research consortia

• National node or EHDS secure processing environment