Bowen Deng Email & Phone Number

South San Francisco, California, US

Santa Monica, California, US

• Design, implement and operate the internal infrastructure for machine learning.

• Led the architecture and implementation of an in-house machine learning platform that streamlined the entire end-to-end process, including data ingestion and pre-processing, model training and optimization, performance.
• Developed and implemented image similarity models for retail items using semi-supervised learning techniques to overcome noisy labels, resulting in a 12% improvement in F1 score compared to competitor's classification.
• Optimized input pipelines for large datasets by leveraging Bigtable, achieving a 2x speedup for model training.

• Designed and implemented rstream, a stream processing library that supports automatic state persistence and failure recovery with reference counting
• Enhanced in-house tracing infrastructure around gRPC and Kafka, making profiling and debugging more accessible across services
• Developed event-sourced microservices for data ingestion and aggregation
• Developed and maintained CI/CD pipelines on Azure DevOps for production services
• Implemented libraries and development tools to support remote debugging for production services (written in Node.js or Python) on Kubernetes, and significantly shorten troubleshooting time for several production issues

• Built a system detecting landing page cloaking (showing reviewers benign pages, but redirecting users to other malicious pages), and the system enforced 130K+ Ads within the first week of its release with a 0.04% false.
• Developed a time-windowed counter service processing 12K log events sampled from mobile clients in real-time and aggregated the events to high-level features for cloaking detection

Redmond, Washington, US

• Developed Cosmos pipelines analyzing large scale (TB-level) semi-structured transactional logs from Microsoft Azure and Office 365, and helped the product teams caught several system failures in their early stage
• Analyzed performance characteristics of distributed jobs for log analysis, identified performance bottlenecks caused by poor data locality and redundant computation, and performed optimization archiving 10x speedup
• Designed an extendable log analysis framework and integrated several existing log-analysis algorithms into the framework

Master’S Degree

Ph.D. (Dropped Out), Computer Science And Engineering

Bachelor’S Degree, Software Engineering