Machine Learning Engineer2021 - PRESENTTikTok
Technologies: C++, Python, Recommendation Systems, Neural Networks
- Improved TikTok's live streaming recommendation system by adding multiple retrieval and recall queues.
- Improved TikTok's live streaming recommendation system by using more complex neural networks and adding more concrete objectives and targets for the system.
- Increased TikTok's overall live-streaming revenue by applying multiple revenue strategies, implementing those strategies in the system, and conducting A/B tests.
Senior Data Scientist2019 - 2020Shopee
Technologies: Algorithms, Anti-fraud, Graph Databases, Apache Airflow, PySpark, eCommerce
- Led a team of five (plus two interns) to build and refactor a unique user identification system, which included setting up a graph database and building a data pipeline and business engine.
- Collaborated with team members to build an interactive data dashboard for managers and other colleagues to explore the user-user relationship.
- Assisted team members with improving JanusGraph, an open-source tool, according to business requirements.
Data Scientist2018 - 2019Shopee
Technologies: Python, Recommendation Systems, Graph Databases, Neural Networks, Deep Neural Networks, Machine Learning, eCommerce
- Increased the click-through rate on the homepage recommender system by 50% by using deep neural networks.
- Increased the conversion rate on the shopping cart page by around 20% by using online learning to build a new module called "You may also like."
- Created graph database clusters and used the database to store buyer-seller relationships.
Machine Learning Engineer2017 - 2017XRVision
Technologies: Python, C++, Computer Vision, Object Detection, Facial Recognition, Machine Learning, Anti-fraud
- Built object detections for CCTV cameras for buildings, factories, and roads, with more than 70% accuracy.
- Developed image classification models for the company's core product that has the general ability to recognize gender, facial hair, eyeglasses, and other features, with an overall accuracy of 90%.
- Built facial recognition models for several companies' check-in systems.