Pittsburgh, Pennsylvania, United States
MS in ECE and BS in ECE + ML Minor from Carnegie Mellon Univerity. Specialized in Multimodal Machine Learning, Explainable AI, and Machine Learning Systems. Researched modality bias in Vision-Language Models at the Human Sensing Lab (paper under review).
- Co-led research with a PhD student on quantifying modality bias in Vision-Language Models. - Designed a novel metric for measuring modality bias and developed the codebase to compute it and run experiments across VLMs. - Co-first-author paper currently under review.
- Co-led a team of 8 teaching assistants for a course of 70+ undergraduates, managing grading timelines and grade calibration to ensure consistent, on-schedule evaluation. - Moderated recitation discussions and collaborated with faculty to revise and improve course assignments based on student and TA feedback. - Discussed AI Ethics and Safety, Big Data Ethics, GDPR and Predictive Privacy in class and recitation.
- Investigated and read state of the art papers regarding token merging algorithms for Vision Models and KV-Caches for Language Models. - Implemented the Heavy Hitter Oracle KV-Cache for the LLaVA Vision-Language Model by directly modifying the LLaMA backbone. - Replicated vanilla LLaVA performance on TextVQA of 61.24\% using a KV-Cache 0.25x the size of LLaMA's context window.
Teaching Assistant for 18-213 Introduction to Computer Systems - Led small group meetings of 4-5 undergraduate and masters students to reinforce concepts from lectures. - Held office hours to provide individual assistance to students on high level concepts for homework. - Covered topics ranging from computer architecture, system engineering, and computer networking. - Co-wrote a programming demo in C to help students understand different types of cybersecurity attacks for an Introductory Systems course.
- Led development of and deployed production code for Delta Robot calibration with a team size of 8. - Produced a program and documentation for Hand-Eye-Calibration for xArm7 Robot Arms, Intel RealSense Cameras and Zivid Cameras using OpenCV. - Researched and documented 23 papers on various Point Cloud Registration for future work with the core product.