• Participated in project of building a chatbot for the AskMargot website using AWS Lex.• Using Lambda function and AWS SageMaker as Machine Learning tools to do sentiment analysis and model training.

• Contributed to the World Model Benchmark Project (Manuscript in preparation), evaluating Large Language Model performances in reasoning• Led data pipeline development and managed data collection using Habitat- Lab and ManiSkill for 500+ cases.• Managed project timelines, task distribution, and documentation.• Led data pipeline development and collected data using Habitat-Lab and ManiSkill for over 500 cases. Utilized 5+ open-source datasets to evaluate models including… Show more • Contributed to the World Model Benchmark Project (Manuscript in preparation), evaluating Large Language Model performances in reasoning• Led data pipeline development and managed data collection using Habitat- Lab and ManiSkill for 500+ cases.• Managed project timelines, task distribution, and documentation.• Led data pipeline development and collected data using Habitat-Lab and ManiSkill for over 500 cases. Utilized 5+ open-source datasets to evaluate models including GPT-4o/4o-mini and Gemini-1.5-flash/pro. Conducted data analysis, data visualization, and drew conclusions. Show less

• Contributed to the project “Evaluation of Programmatic Cycle Consistency of Large Language Models” (workshop manuscript in preparation)• Created a dataset of 20,000+ images using Stable Diffusion and conducted over 20,000 experiments on models including CLIP, SigLIP, Llava, PaliGemma, and Llama.• Identified inconsistencies in generation and comprehension, and assessed ViperGPT for programmatic inconsistency analysis.

• Project: Enhancing Speech Recognition using Machine Learning Tools • Explored AI bias in speech recognition tools (e.g., Montreal Forced Aligner, Char-siu, Whisper) for diverse speech types-standard American English, accented speech (Mandarin, Spanish), child speech, and distorted speech (noise-vocoded, sine-wave)-to improve inclusivity and accuracy.• Responsible for data pipelines using Whisper, ran 40000+ experiments. Responsible for data analysis using R and paper writing.

• Conduct experiments aimed at demonstrating rats’ ability to transition visual stimuli from an overhead perspective to an eye-level perspective.• Responsible for data analysis using DeepLabCut and SLEAP, created machine learning models for animal behavior video data processing to gain positions of rats for data analysis, analyzed 200+ minutes videos.• Participate in drive building session of brain-computer interaction imposing on rats.

• Participated in AI/ML research to automatically score interview videos, developing and optimizing machine learning software solutions.• Developed product-focused AI solutions using computer vision and natural language processing technologies.

• Project: KidTalk and L2Talk• Explore the gap between monolingual children's language output and language perception (KidTalk), and the gap between bilingual adults’ language output and language perception (L2Talk).• Collected 72 3–5-year-old child word-naming experiment data, free speech data, and Goldman-Fristoe Test of Articulation (GFTA) data using MATLAB, Phon, Praat, and Eyelink1000; contributed to experiment design, collected 5 pairs adult data of word-naming; participate in… Show more • Project: KidTalk and L2Talk• Explore the gap between monolingual children's language output and language perception (KidTalk), and the gap between bilingual adults’ language output and language perception (L2Talk).• Collected 72 3–5-year-old child word-naming experiment data, free speech data, and Goldman-Fristoe Test of Articulation (GFTA) data using MATLAB, Phon, Praat, and Eyelink1000; contributed to experiment design, collected 5 pairs adult data of word-naming; participate in data analysis using R for both initiatives. Show less