Deliverable 2.11 on “Federated Infrastructure Design” presents the results from M1 to M18 of Task 2.6, which will continue until M42. In this deliverable, we report the design of the Federated Learning (FL) infrastructure and the tests we conducted. The current federated infrastructure is: secure thanks to the encrypted communication via gRPC/TLS and secure aggregation (SecAgg/SecAgg+); flexible, because it supports horizontal (same data types) and vertical (different data modalities) FL workflows, including embedding-based vertical FL; and, deployable, because we containerised via Docker and reproducible across HPC (Slurm), cloud (Azure), and manual setups. These features were validated and refined through two project workshops: Workshop 3 (Jan 30 & 31 2025), which focused on horizontal FL with XGBoost and CNNs across local, cloud, and multi-site environments. Workshop 4 (May 19 2025), which focused on vertical FL across three distributed data sources (i.e., SURF, UNIPD, and UNITO) using split ALS data modalities. We have identified a minimal participant configuration that allows access to a server provisioned with 16 vCPUs and 32 GB of RAM; clients can optionally leverage GPUs. In terms of scaling and monitoring, we are using Flower Next with persistent servers, CLI, and Messaging API; basic logging is implemented, and we plan to integrate MLflow/W&B. From a privacy and security perspective our approach combines TLS and SecAgg and we are planning upgrades to Differential Privacy (DP), Trusted Execution Environments (TEEs), and defenses against poisoning. Additionally, we tested our data approach on the BrainLat dataset (MRI + EEG), though limited by low matched-patient count and data errors, the embedding architecture remains adaptable. This deliverable verifies Milestone 6, “Data management infrastructure”, by testing the Federated Data Management Infrastructure on use-cases 1 and 2, including data and involved partners.