Efficient & Adaptation Methods

We make large models practical — parameter-efficient fine-tuning (LoRA variants), knowledge distillation, pruning and quantization, and transfer, meta, and continual learning that adapt foundation models to new domains at a fraction of the training cost.