Research Areas

Agents that improve through governed experience without human intervention in the loop.

Learning continuously without catastrophic forgetting, with memory consolidation and decay.

How memory quality, retrieval, and consolidation change with experience and scale.

Emergent capabilities from skill composition, validation, and iterative improvement.

Teacher-student distillation, LoRA/PEFT adapters, quantization-aware distillation.

Multi-armed bandit and policy learning for optimal model selection.

Models that adapt at inference time without gradient updates.

Genetic algorithms, neuroevolution, quality-diversity for neural architecture search.