Mountain View, California, United States
AWS Redshift Data Management
AQUA is a distributed and hardware-accelerated cache to perform near-storage computation for AWS Redshift. - Gathered metric requirements from the hardware team and data plane team, and composed a list of 25 metrics needed. - Designed and built the per-request structured metrics and dashboards for the hardware management layer to simplify the debugging and bottleneck tracking process, with the capacity to handle around 1,800 requests per second. - Discovered the log missing issue for AQUA in production and provided mitigation methods for the data plane team.
- Constructed and maintained a local developing environment (Docker image) with Online Learning SDK for internal online learning platform users, smoothing the development and testing procedures. - Constructed an embedded module to monitor positive and negative training samples in Java. - Designed and implemented a real-time data visualization module to display feature distributions during the training pipeline in Java and Python, facilitating the development and debugging process for users.