Congratulations to our labmates, Amir Mohammadi and Jarin Ritu, on passing their preliminary exam and becoming PhD candidates! Amir’s research develops a distribution-guided parameter-efficient fine-tuning (PEFT) approach that integrates histogram-based statistical adaptation to more efficiently and robustly fine-tune foundation models for waveform-based sensing (e.g., passive sonar). Jarin’s work introduces a texture-aware knowledge distillation framework that enables compact, smaller (“student”) models to retain fine-grained detail and improve performance across different tasks such as environmental sound classification and cell microscopy segmentation.
