Jun Wang
Development
MA, United States
Skills
Machine Learning (ML)
About
JUN WANG's skills align with IT R&D Professionals (Information and Communication Technology). JUN also has skills associated with Programmers (Information and Communication Technology). JUN WANG has 7 years of work experience.
View more
Work Experience
Deep Learning Engineer (Research Intern)
UII America, Inc
November 2022 - May 2023
- Pioneered an innovative human pose estimation pipeline integrating bottom-up and top-down methods for addressing complex occlusions in diverse environments. • Authored 15K+ lines of code from scratch to build the network, employing various skills in software development and deep learning/CV tools, including PyTorch and OpenCV • Conducted comprehensive data augmentation and preprocessing for the training on the 18GB COCO dataset (118K images), achieving competitive performance metrics (Test AP: 0.40, AR: 0.75) on OCHuman dataset, closely rivaling state-of-the-art models in human pose estimation.
Research Assistant
Rensselaer Polytechnic Institute
August 2016 - August 2022
- Developed algorithms for complex decision-making systems in voting theory, applying innovative search methods and machine learning strategies to reduce runtime and improve efficiency in analyzing large datasets. • Designed and implemented advanced neural network models, resolving data insufficiency issues through a novel multi-model training framework, and utilized Gurobi and IBM ® ILOG CPLEX for constrained optimization subproblems. • Achieved notable efficiency improvements across various topics, including an average 14.28% reduction in runtime and a 67.01% decrease in time to result convergence. Part of this work was published at the AAAI conference, known for its competitive 16.2% acceptance rate.
Machine Learning Research Associate
IBM Research
May 2020 - August 2020
- • Investigated out-of-distribution (OOD) generalization in machine learning models, with a focus on nonlinear and polynomial models, achieving key insights in robust model prediction. • Pioneered approaches in handling distributional shifts involving confounders and anti-causal variables, enhancing model adaptability and accuracy in diverse environments and complex scenarios. • Contributed to an ICLR 2021 publication (28.7% acceptance rate), providing new perspectives on Empirical Risk Minimization (ERM) and Invariant Risk Minimization (IRM), and their implications for model generalizability.
Machine Learning Engineer
Alibaba Group
June 2019 - August 2019
- Developed a novel ranking algorithm to balance revenue, welfare, and user utility, effectively navigating the trade-offs among these objectives. Conducted extensive tests using synthetic datasets to validate the algorithm's performance and optimize its parameters. • Implemented strategies for achieving a tighter bound on Revenue Welfare under normal distribution assumptions, utilizing contour plots for visual assessment and impact analysis.
Machine Learning Engineer
EMSQRD LLC
July 2023 - Present
- Engineered and deployed an AI-driven calibration system capable of real-time updates for competitive racing analysis, processing over 140K races and 100+ features, employing advanced models like Random Forest, GBM, and Neural Networks using PyTorch and Scikit-learn. • Enhanced predictive accuracy by integrating multi-faceted data sources into model training, optimizing model performance through extensive hyperparameter tuning and feature engineering, achieving a 53.6% MAE reduction over baseline AI predictions. • Visualized model performance and calibration using custom-plotted figures, illustrating error distribution and statistical analysis, which aided in model interpretation and validation.
Education
Rensselaer Polytechnic Institute
Ph.D. in Computer Science
August 2016 - August 2022
Rochester Institute of Technology
