I work on XBorg which manages Google GPU/TPU fleet, orchestrates and schedules Google‘s AI/ML workload. Built the cut-edge training infrastructure for LLM.

- Designed and built the durable storage infrastructure for Blobstore (GCS), a planet-scale storage system.- Driven AI initiative with 4 engineers to build AI agent for on-call ticket handling, including content summarization and action item generation. - Led 5+ engineers to deliver the next generation of planning engine resolving how to efficiently place bytes across Google's global data center. - Designed and Implemented multi-step execution and hand-over mechanism for Blobstore data placement across regions, improving the efficiency by 30%.- Defined and implemented the metro rebalancing SLO, which quantified the rebalancing performance at the metro level and achieved a peak data move rate at 50 Tib/s.- Implemented a safety protection mechanism to prevent the storage system from reducing durability due to unintended hardware operation.

built the next-generation equity trading platform which serves tens of millions of retail investors and brings the best trading experience to retail investors.

- Founding Engineer on Porting Assistant For .NET - an open-source code analysis and translation tool for .NET.- Led the open-source effort of Porting Assistant for .NET tool, and redesigned the whole backend architecture and external APIs.- Designed and implemented the telemetry data pipeline, which collected metrics and logs from Porting Assistant tool, uploaded and processed through API Gateway, S3, SQS, Lambda function, and Elastic Search.- Designed and implemented a distributed data pipeline that runs to collect 2 million NuGet packages, analyzed by Lambda function, and generated compatibility benchmark stored in Dynamo and Elastic Search.

- Built the Control Plane of Riverbed's SDN Product: SteelConnect.- Features worked on included system-dump service, Redis database, ETCD cluster whack, HA failover, and preferred mastership.- UI development for the whole network routing team and CDL orchestration experience with DHCP, Multicast and Loopback IP, etc.

- Detect the network attack based on the end-point traffic metrics.- Built simulation tools for large scale and volume Shrew Denial of Service attack based on Mininet.- Deployed sampling algorithm to analyze the spectral pattern of network traffic with cross-over trace.- Investigated the issue of tuning DoS parameter and trade off between attack traffic rate and synchronization