Graded assignments and homework for the course COMPSCI 348: Principles of Data Science taught by Prof. David Jensen

• Assisted undergraduate and graduate students with various stages of internship and job search

• Worked with students on networking strategies, resume reviews, and interview prep for software-centric and data science-related roles

• Aimed at evaluating the adversarial robustness of explanation methods for NLP model using saliency maps and integrated gradients explanation method utilizing techniques used in image domain

• Evaluated RoBERTa base model on word deletion, inflection, synonym substitution and misspelling attack using a novel greedy based attack strategy to achieve successful attack rate in the range of 32-86 %

• Extended Box Embeddings library to TensorFlow, designing custom layers to perform numerically stable geometric operations, including their smoothed variants and custom regularization

• Wrote unit tests (≥86 % coverage) ensuring TensorFlow layers matched their PyTorch counterparts, and merging the code into a unified codebase with existing continuous integration

• First author (joint) on the paper ‘Box Embeddings: An open-source library for representation learning using geometric structures’ [EMNLP 2021]