Jingyi Gu Email & Phone Number

New York, Ny, Us

Princeton, New Jersey, United States

Finetuning and implementing VLMs/LLMs on images for object extraction, code generation, and Q&A assistant

Tarrytown, New York, United States

- Generate data-driven machine learning framework to gain insights in human biology and disease.

San Jose, California, United States

- Explore pre-train models under self-supervised learning paradigm for molecular representation learning- Utilize GNN and Transformer as backbones to explore 2D topology and 3D spatial position of molecules– Finetuned the molecular representation for downstream deployment (e.g., property prediction)

Woodbridge, Nj

- Tested impact of vehicle information on predicting loss in auto insurance to optimize pricing model, automated end-to-end process, improved model performance by 5%- Implemented tree-based model structure on pricing model, delivered robust ETL and modeling process.- Refreshed logistic regression model to predict policy lifetime expectancy, monitored policy retention trend over time to allow early detection of trend change- Clustered GPS coordinates in NJ crash accident records in 2006-2016 to territory level by K-Means, detected dangerous areas with high crash rate, visualized crash location distribution on the map-

Piscataway, Nj

- Developed stochastic deterioration model in R based on survival analysis, Markov Chain to forecast survival performance of 150k Mid-Atlantic bridge components, visualized simulation results on bridges condition rating, provided recommendations and insights- Designed ETL process capable of bridge data migration from LTBP data warehouse to AWS S3 platform- Conducted machine learning and statistics algorithms on massive bridge data to implement dimension reduction in R, which improved 20% of efficiency

Shanghai

- Generated data processing pipelines on a clinical trial study data (100k+ rows) to detect discrepancy between drug administrating rules and protocol in R, guaranteed safety results- Automated a report dashboard for clinical trial study data integration by R Shiny, improved efficiency by 10 times- Supported in coding training for entry level statistical programmers, primarily focus on R packages (dplyr, lubridate, ggplot2, stringr)

Doctor Of Philosophy - Phd

Master Of Science

Bachelor'S Degree

Education record