PhD Student in Computer Science at Rutgers University, applying large language models and machine learning to creative design and pedestrian modeling.
We built a domain-specific knowledge graph for architecture design and integrated it with LLMs to assist early-stage design question answering.
Given a textual card game description, Cardiverse uses an LLM-powered framework to create novel and playable game variant prototypes.
Utilizing Large Language Models as an organizer of various generative tools, the StoryAgent framework automates digital storytelling with fine-grained control over intermediate products.
How will attention affect walking speed in retail areas? We model how pedestrians will be attracted to environmental objects and how this will slow down their walking speed. Then we show how our model can help optimize the design of retail areas in transportation hubs.
Search for architectural case studies using natural language queries, images, or just a few mouse clicks.
We form the trajectory prediction problem as a denoising impaint task and design a map-based guidance term for the diffusion process. It generates trajectory predictions that adheres almost perfectly to environmental constraints.
A synthetic data generator for multi-view multi-group multi-person human atomic actions and group activities, which facilitates the learning of human-centered tasks across single-person, multi-person, and multi-group conditions.
The influence of arriving passengers and the interaction effects between external factors are compared in two scenarios: the mutual recognition of security checks and the application of face recognition systems.
A detailed empirical observations that focus on pedestrian speed variations and their dynamics in front of stores.
Which parameters are most sensitive, thus most important to calibrate, in pedestrian simulation models on train station platforms? We conduct a sensitivity analysis to answer this question.
