Ky-Vinh Mai

NLP

Language Science Department - Student Researcher under Dr. Lisa Pearl's Computational Language Lab and Dr. Greg Scontras' Meaning Lab. - Developed search functions to identify ambiguous sentences with particular dependency structures, consistently achieving accuracy rates ranging from 94% to 100% on test data. - Developed an ELT pipeline for large scale corpus and speech analysis. Used SpaCy, Selenium, BeautifulSoup, and PyTorch to dynamically search and scrape web pages for transcripts and audio. Generated a multimodal dataset with thousands of text and audio recordings, segmenting and force aligning the text and audio, using OpenAI Whisper and CTC algorithm Skills: Python (Programming Language) · Spacy (Natural Language Toolkit) · Machine Learning

Lead undergraduate researcher working in a 10-week research program on the Intonation of Scope Ambiguity project. I work in a team of 3 students in the Ambiguity Lab.

- Constructed a part of speech tagging algorithm and performed corpus analysis for endangered language isolate Purépecha. - Developed novel POS tagging algorithm for Purépecha (endangered language isolate, 100k+ speakers) - Implemented Helsinki FST-based morphological segmentation system - Standardized orthography system to support computational analysis

Was one of 9/500 student researchers selected. Advised by Melanie Mitchell, the research project investigated conceptual analogy making abilities in large language models. - Created a pipeline that finetunes, tests, and refines OpenAI GPT models, under a strict budget ($500). - Implemented different prompt engineering techniques. Discovered a performance increase on the Conceptual Abstracting Reasoning Corpus (Concept-ARC). - Conducted statistical analysis and T-SNE visualizations on model embeddings. - Presented these findings to esteemed researchers at the institute (including my mentor, Melanie).

Investigating Bias in Foundation Models

- VP of External Affairs and Co-Founder - Created and oversaw a board of 9 members to grow a thriving and supporting club community - Lead outreach to external organizations for professional panels and networking, workshops, and social events attended by up to 405+ undergraduates

A mentorship program for inspiring STEM first-generation high-school students, to help assist them in the academic process and future careers. - Created workshops for STEM topics and mentored a group of students throughout their summer program: assisting with essays, mathematics, projects, etc.