Assess students' ability to design a practical privacy‑preserving federated learning system for resource‑constrained mobile clients that supports heterogeneous client models; require selection and integration of differential privacy mechanisms (privacy accounting, noise calibration), secure aggregation protocols, and communication compression techniques (quantization, sparsification, sketching); include an architectural diagram, protocol flow, algorithm choices (e.g., FedAvg/FedProx, personalization strategies), analysis of utility–privacy–communication–compute tradeoffs, a threat model with mitigations, and an evaluation plan with metrics and experiments for mobile deployment.