bilibili
Bilibili (stylized in all lowercase), nicknamed B Site, is a video-sharing website based in Shanghai where users can submit, view, and add overlaid commentary on videos...
As a passionate Machine Learning Engineer, I specialize in crafting advanced recommender systems that captivate millions by suggesting videos tailored to their unique interests.
My commitment to excellence in machine learning and my continuous drive to stay at the forefront of technological advancements enable me to deliver impactful and innovative solutions that enhance user experiences.
My expertise spans the entire lifecycle of recommender systems, including user profiling, model training, and seamless model deployment. My journey includes traditional statistical methods like ItemCF and Swing. Embracing the embedding-based retrieval system, I have leveraged algorithms such as item2vec and pretrained embeddings like BERT and CLIP for sophisticated item2item retrieval, enabling me to create highly relevant and personalized recommendations.
Recently, my focus has been on training a Deep Structured Semantic Model (DSSM) for refined user2item retrieval. This model has substantially improved our recommendations, leading to a more engaging user experience. Additionally, I have delved into the intricacies of complex ranking models to gain deeper insights into user preferences and behaviors. This endeavor has led to a substantial improvement in Click-Through Rate (CTR), highlighting the effectiveness of my models in understanding and anticipating user needs.
- Deep Structured Semantic Model (DSSM)
- Wide & Deep Models
- DeepFM
- Deep Interest Network (DIN)
- Product-based Neural Networks (PNN)
- Deep Interest Network (DIN)
- Squeeze-and-Excitation Network (SENet)
My commitment to excellence in machine learning and my continuous drive to stay at the forefront of technological advancements enable me to deliver impactful and innovative solutions that enhance user experiences.