The project aims to enhance multimodal question-answering capabilities by developing a model that programmatically integrates textual and visual information to generate accurate, context-aware responses.
Improving LLM's multi-token prediction accuracy with cascading connection scheme among the decoding heads.
A web app that allows you to chat with YouTube video contents with references, powered by GPT3.5 and deployed on AWS cloud.
A swarm of Ummanned Surface Vehicles for collecting Real-time maritime data.
An Intelligent Tutoring System (ITS) to Train Psychomotor Skills in Virtual Reality.
With the increasing complexity and overall scale of chip design in the post-Moore’s-law era, chip packaging optimization has become an important issue with many design variables and objectives. I led two other undergraduates and proposed applying Bayesian optimization machine learning algorithms to tackle this problem.
A mobile robot solution for manure pollution caused by animal husbandry.