Machine learning scientist at Expedia Group

Seattle, WA

​Jan. 2020 - now

 

Apply machine learning methods such as online convex optimization, Actor-Critic algorithm, tree algorithm, and neural network to improve kpis in online ranking problems as well as bidding of search engine market. 

​

Intern at 3M

St. Paul, MN

​May 2018 - Aug. 2018

 

Project: Deep Learning-based time series data classification​

• Generated synthetic power network data by Matlab/Simulink with actual power consumptions.

• Preprocessed the data and reformulate the power network anomaly detection and localization problem as a time series data classification problem.

• Applied deep learning tools including CNN and LSTM with different setups to extract both the spatial and temporal relations of the data under the tensorflow environment written by python.

• Fine-tuned the hyper-parameters of the proposed deep learning algorithms to achieve more than 94% accuracy with only 20% of the sensors.

 

Intern at Boston Scientific

Arden Hills, MN

​May 2017 - Aug. 2017

 

Project 1: Data analysis on the demographic data from the Heart Failure patients

• Applied hierarchical clustering algorithm to obtain different sub-groups, whose variables were grouped and extracted by leading eigenvectors, and the clustering was evaluated by B_k statistic and Cubic Clustering criterion.

• Carried on the significance analysis for every variable among different groups. Related each sub-group’s heart failure event rate and different sensor signals with the group’s characteristic description to investigate the influence of the demographic data.

​

Project 2: Adaptive hyper-parameter optimization by reinforcement learning

• Designed the reward function based on the performance of the HeartLogic device.

• Applied the multi-armed bandit algorithm to automatically tune the hyper-parameter in HeartLogic device based on the information the patients/doctors provide with retrospectively.