Projects
Selected research projects in federated & continual learning
CRAFT
CRAFT
Forgetting-aware, intervention-based adaptation that lets continual learners absorb new tasks without overwriting prior knowledge.
Continual Learning
LLMs
Read more →
SCOPE
SCOPE
Communication-efficient federated learning that builds semantic coresets via orthogonal projection embeddings to tame class imbalance.
Federated Learning
Coresets
Read more →
FedReFT
FedReFT
Federated representation fine-tuning with an "All-But-Me" aggregation scheme — adapting hidden representations without sharing full models.
Federated Learning
PEFT
Read more →
Fair BW
Fair Bandwidth Allocation
A game-theoretic model and distributed heuristics for fair edge bandwidth allocation across concurrent federated learning processes.
Game Theory
Edge Computing
Read more →