# Federated AI

The collaborative-learning layer: federated learning, privacy and evaluation.

- [Federated learning](federated-learning.md) — Many clients train a shared model under orchestration while data stays decentralised.
- [Secure aggregation](secure-aggregation.md) — The server learns the aggregate of updates, not any individual update.
- [Cross-device vs cross-silo](cross-device-vs-cross-silo.md) — Large fleets of weak devices vs a few institutions with legal separation.
- [Federated analytics](federated-analytics.md) — Distributed statistics and monitoring without training a model.
- [Federated evaluation](federated-evaluation.md) — Measuring models when data and outcome definitions stay distributed.
- [Non-IID data](non-iid-data.md) — Heterogeneity is intrinsic, not incidental, in federated settings.
- [Differential privacy](differential-privacy.md) — A formal bound on what any individual contributes to a released result.
- [Trusted execution environments (TEEs)](trusted-execution-environments.md) — Hardware-isolated computation strengthening server-side confidentiality.
- [Homomorphic encryption & MPC](homomorphic-encryption-mpc.md) — Computing over encrypted data / across parties without revealing inputs.
- [Membership inference & gradient leakage](membership-inference.md) — Privacy attacks that recover information from model updates.
- [Model poisoning & backdoors](model-poisoning-backdoors.md) — Malicious participants corrupt the shared model.
- [Federated unlearning](federated-unlearning.md) — Removing a participant's influence from a trained model.
- [Federated / confidential RAG](federated-rag.md) — Applying the federated pattern to LLM-era retrieval and generation.
- [Multi-dimensional evaluation rubric](evaluation-rubric.md) — Judge federations on utility *and* governance, not accuracy alone.
- [Hub-and-spoke vs hierarchical coordination](hub-and-spoke-vs-hierarchical.md) — Centralised aggregation vs hierarchical/asynchronous/enclave-backed designs.