• Developing AI solutions by utilizing LLMs and embedding models in the market• Developing general purpose embeddings from both structured and unstructured data for applications such as downstream modeling, recommender system and clustering/segmentation.• Developed a deep-learning based pricing model to price the function rooms (e,g. conference rooms ormeeting rooms) for all Hilton hotels in the world.• Advising leadership on cloud infrastructure (AWS) setup and data science application production pipeline• In charge of developing internal data science tools/library (e.g AWS, deep learning, data loading and transformation) to help data scientist simplify and streamline their daily workflow and modeling process.• Mentor junior data science members and host knowledge sharing sessions to inform data scientists on the latest development in machine learning.

- Science lead for the ML-based demographics (e.g age, gender, marital status etc) prediction models that are used by Amazon teams to identify underserved segments of population. Improved the model performance by 5% to 40% (depending on types of demographics models) over the previous implementations.- Initiated and developed the team's first shared code library to allow rapid development of machine learning models, pipelines and data science applications. The tools include python functions and classes to simply and speed up many of the teams daily tasks (AWS services such as S3 and EMR, ML model training, distributed computing, data-loading).- Implemented the team's first deep learning capability using PyTorch, including distributed training to speed up large scale training. Provides tools to train simple multilayer perceptron neural net models rapidly (GPU and/or distributed). Also provided frameworks for team member to rapidly develop more complicated neural net structures and trainers.- Implemented a custom out-of-core data-loading solution for batch training (both single node and distributed) machine learning models using large dataset (100+GB). The solution allows fast shuffling for large data, fast data splitting for cross-validation purposes and near real-time data streaming.- Automated data pipelines using AWS CloudFormation to allow rapid deployment (or removal) of machine learning applications for multiple marketplaces.

In-house data scientist for Capital One's Spring product, which provided best-in-class discount deals curated by the Spring team for small business customers.- Developed and maintain a matrix-factorization based recommender system for providing personalized deal suggestions- Perform marketing campaign segmentation using machine learning and data to identify the right target customers- Construct data tools and data pipeline to generate clean and enhanced datasets. Help the team produce better insights and accurate reports.- Automate periodic reports for data quality, Spring product usage and marketing campaign performances

• Design a solution recommendation system that automatically suggests procedures relevant to diagnose/repair clients’ machines based on machine sensor data input. Work closely with product and engineering teams to ensure implementation is error-free and efficient. Regularly seek feedbacks from clients to validate results, improve the quality of recommendations and address any outstanding issues.• Devise a general solution to detect anomalous event patterns for use cases such as detecting new types of machine anomalies and anomaly detection in network security. Collaborate with product, engineering and user experience team to explore additional use cases, shape product direction and understand ease of implementation of various data science solutions.• Developed an antibiotic recommendation system, which predicts the effectiveness of various antibiotics for individual patients, to help doctors prescribe treatments for patients with urinary tract infection. Based on test data, our system increased the effectiveness of treatment by 18%.• Developed a data science model to help hospital client recover missing revenues. Our model identifies potentially missing charges by suggesting medical services/items that are often associated with a particular procedure but are missing from the bill.

- Representation. Spokesperson for the organization in local and international events (e.g. COP15 conference).- Program management. Initiated and implemented the foundation’s first grass-root energy and climate change education program, which became the template for other education programs in the organization.- Energy policy research. Assisted a congresswoman to draft Taiwanese Climate Change Act by researching energy policies of foreign countries.

- Scientific Research. Conducted x-ray experiments to study dynamics of polymer thin films and analyze acquired experimental data.