Work Experience

@MRS AI Foundation Model Co-design Team

Benchmarked and fine-tuned multiple state-of-art OCR vision language models (vllms)
Generated 10 million synthetic vision question-answering pairs for fine-tuning vllms

Decomposing d3.js: Behavior-Driven Testing of Big Data Exploration Tools

Collaborator: Professor Leilani Battle, Paul G. Allen School of Computer Science & Engineering

Designed a collection of 50 D3 visualization introduction tutorials using various climate change datasets
Train machine learning models and/or LLMs to generate customized learning content for D3 users

Systems and Modeling for Accelerated Research in Transportation Project

Sponsor: Argonne National Lab, US Dept. of Energy

Designed blocked factorial experiments and deployed an online survey application collecting major
ridehailing driver behaviors on AWS
Explored prospective effects of trip-level characteristics on autonomous vehicle ownership choices in the US by developing an integrated choice and latent variable (ICLV) modeling framework
Investigated the impacts of trip features on ride-hailing drivers' response to trip requests using a generalized additive mixed model (GAMM)
Jointly modeled ride-hailing drivers’ working time and relocation preferences by developing a mixed logit discrete choice model framework

Promises of Data from Emerging Technologies for Transportation Applications

Sponsor: US Federal Highway Administration

Processed over 700 million rows of individual trajectory records from over 460k mobile phone users of Puget Sound region.
Investigating the causal relationship between spatial/temporal properties of mobile phone trajectory big data and derived human mobility patterns.
Investigated big data quality issues of both passively and actively solicited data during data collection, processing, and analysis.

Examining Health Impacts of Built-Environment in China's Transit Metropolis: A Physical Activity Perspective

Sponsor: National Natural Science Foundation of China

Investigated the relationship between the use of public open spaces and built environment factors, time-varying factors, and socio-demographics using a Generalized Linear Mixed Model (GLMM) model

Promoting Built Environment Quality for Cycling Using Dockless Public Bike Data

Sponsor: National Natural Science Foundation of China

Investigated the relationship between dockless bike usage and the built environment in Shanghai using a Generalized Additive Mixed Model (GAMM)

Instructor: Professor Ana Pinto da Silva, Human Computer Interaction + Design

Co-designed course contents, graded student assignments and assisted class activities
Gave a guest lecture on “Power of Data Visualization”

Instructors: Prof Leilani Battle, Jon Froehlich, Tal Wolman, School of Computer Science & Engineering

Instructor: Professor Cynthia Chen, Civil and Environmental Engineering

Designed 4 “Python tutorial” Google Colab notebooks, 4 homework assignments, and 3 course projects
Gave 2 Python tutorial lectures