Sneha Rao

Conducted in-depth literature reviews on corrective machine unlearning to identify gaps in current approaches. Worked under Prof. Elliot Creager to design novel algorithms addressing real-world data deletion demands in machine learning.

Conducting an in-depth literature review on corrective machine unlearning to identify gaps and limitations in existing approaches. Working under the guidance of Prof. Elliot Creager to design a novel algorithm addressing corrective machine unlearning in response to real-world data deletion demands.

Developed proof of concept for crack segmentation in sealant images using Faster R-CNN, Unet, and Segformer models with a Dice score of 0.74. Performed image preprocessing, data augmentation, and analysis of climate data to explore environmental correlations with crack length. Analyzed NRC IRAP data to extract research trends using keyword extraction techniques and constructed a large dataset of AI publications for topic modeling and weak signal analysis, utilizing GPT-4 for interpretation.