Rex Wang Email & Phone Number

Seattle, Wa, Us

Cupertino, California, United States

Cupertino, California, United States

1. Created a localized development cloud environment (10+ microservices) that mimics the Mist Cloud architecture to streamline developers’ incremental development cycle2. Redesigned Redis, Postgres DB, and Kafka’s data pipelines and streaming topologies to separate the local dev cloud completely from the staging/production environments3. Virtualized Juniper’s Ethernet switch behaviors in the dev cloud to save up hardware and energy consumption

Beijing, China

• Re-architected (localized) IBM’s Product Optimization (PO) ML service using open-source components including Ray Serve, MinIO, JupyterHub, etc. in order to let local customers purchase PO service at a lower price• Embedded PO scoring service on top of Ray Serve with FastAPI framework for frontend HTTP request• Optimized the deployment process of customer’s ML scoring service from local Helm Chart to online http request• Automated the entire deployment process of PO on Kubernetes… Show more • Re-architected (localized) IBM’s Product Optimization (PO) ML service using open-source components including Ray Serve, MinIO, JupyterHub, etc. in order to let local customers purchase PO service at a lower price• Embedded PO scoring service on top of Ray Serve with FastAPI framework for frontend HTTP request• Optimized the deployment process of customer’s ML scoring service from local Helm Chart to online http request• Automated the entire deployment process of PO on Kubernetes (K8s) using Docker and Helm Charts Show less

Chicago, Illinois, United States

EVSTORE: Scaling Embedding Tables for Deep Recommendation Systems (ASPLOS’23 Accepted)• Created 3 EV-based cache replacement algorithms for increasing memory usage of large recommendation models• Implemented these algorithms using C++ and Python with a result in improving perfect hit in real-world storage workload by up to 400% compared to other policies• Implemented these algorithms with multilevel mix precision caching to Meta’s DLRM and yielding up to 23% and 27% reduction on the… Show more EVSTORE: Scaling Embedding Tables for Deep Recommendation Systems (ASPLOS’23 Accepted)• Created 3 EV-based cache replacement algorithms for increasing memory usage of large recommendation models• Implemented these algorithms using C++ and Python with a result in improving perfect hit in real-world storage workload by up to 400% compared to other policies• Implemented these algorithms with multilevel mix precision caching to Meta’s DLRM and yielding up to 23% and 27% reduction on the average and p90 latency while quadrupling throughput at 0.2% loss in accuracy Show less

Beijing, China

• Processed raw data from hospital’s databases to formatted, washed data for future modeling using Python• Utilized Regular Expression to generate information from medical data to meet 95% accuracy requirement

Master Of Science - Ms, Information Networking

Selected courses: Web Development, Distributed Systems, Cloud Infrastructure, Storage Systems, Mobile and Embedded Software Design, Intro.

Bachelor Of Science - Bs, Computer Science, 3.96

Selected courses: Data Structures and Algorithms, Computer Network, Operating System, Computer Architecture, Principles of Database.