- Experience with event-driven processing using Kinesis, SQS, and lambda - Design and development of scalable and flexible EDM services for ecommerce, with a focus on high tolerance and reliability - Proficiency in using serverless frameworks to accelerate development - Maintained multiple lanaguages application, such as type script, ruby on rails...etc - Ability to introduce and educate team members on new technologies and best practices

- Researched, designed, and built machine learning systems for customer life time value prediction, product recommendation. - Enhanced overall search performance using a reinforcement learning algorithm (PPO) to fine-tune different weighted factors. - Executed key phases of the ML lifecycle including data extraction, exploratory data analysis, data preparation, model development and training, model evaluation and validation, and model serving - Increased recommendation system performance by introducing a microservice framework and refactoring the system, resulting in a throughput of 1k/sec.

Atlanta, US Implemented NLP - SQL models, which allow transform queries and Table scheme into SQL. Tested different DL framework covered pointer network, attention, reinforcement learning and its hybrid. Entities recognition based on database knowledge extracted the relation depends on the corpus. Implemented rule-base NLG model transform SQL result into conversation with flask wrapper.

Atlanta, US Analyzed the correlation between customer behavior and product description for optimizing click-through rate. Built a model to help the search engine to guess unseen search keyword and give a possible estimation. Evaluated value of additional content, such as user review, unconstructed product description. Trained industry-specific word embedding based on Home improvement forum content for solving synonyms that increased performance by 3%.

Chicago, US Predicted airport check-in waiting time through information extracted from tweets. Crafted pre-process pipeline, integrated Tokenizing, Normalization, Lemmatization, Embedding. Designed and developed schema of NLP models linear regression, Xgboost .

Taiwan Interpreted feature engineer on time-series data with Kalman-filter and rule-based. Identified gym excises with user independent classifier. Ported matlab algorithm into C for arm-based MCU.