Shizhen Wang

Lead Zoom unified AI infrastructure and framework to drive all functionality, capacity, scalability, and reliability needs for Zoom’s AI training, inference, deployment, and operation

Working on speech recognition

Working on dialog system and its applications in intelligent customer service, assistance, conversational recommendations, etc.

Working on distributed machine learning on cluster computing infrastructure.

Metanautix was acquired by Microsoft in Dec, 2015. Worked on big data query engine core design and development, including: 1. Machine-learning based query optimizer and query scheduler; 2. Connection to various data sources: Azure SQL, HDInsight, MongoDB and Hadoop HDFS; 3. Support for composite (nested) and array (repeated) data types; 4. Efficient key space sampling algorithm for database partition

Worked on improving Bing Local Search relevance - Discriminative semantic similarity features - Distance normalization model - Business density features - Joint system optimization Worked on Bing Suggests answer services to provide personalized recommendations for places, events, activities, etc.

Worked on normalization and adaptation to improve speaker robustness and noise robustness of continuous speech recognition - Kid's speech recognition - Online vocal tract length normalization (VTLN) for robust speech recognition - Adaptive Histogram Equalization (HEQ)