Colin (Xinyu) Yang Email & Phone Number

California, United States

Sunnyvale, Ca, Us

Observability team

Los Angeles, California, Us

Improve the performance and reliability of backend Python Django code of www.bruinwalk.com, a professor/apartment review sharing platform for UCLA students.Web Application Performance: Rewrote legacy (1 year) Elasticsearch query code with a persistent connection pooling library, and improved performance, readability and reliability of Python Django code around the site. The website can now handle the peak traffic that it previously crashed under.DevOps Practice: Set up production and cross platform (Windows, Mac, Linux) development environment using Docker and Docker-Compose. Led and troubleshooted developers’ migrations to Docker-based dev environment with minimal interruption to their workflow. Improved deployment tooling with bash scripts and Python.

Java Code Safety through JVM Bytecode Instrumentation: https://github.com/colinxy/jrewriterDesigned a Java ClassLoader that improves safety of Java code with minimal performance impact (< 10%). Utilized Java Native API and Javassist compiler API to build a Java ClassLoader that programmatically modifies JVM bytecode sequences at load time to guarantee atomicity in counter increments (i++) with compare-and-swap instructions. Advised by Todd Millstein (http://web.cs.ucla.edu/~todd/).

Sunnyvale, Ca, Us

API Rate Limiting: Designed and developed API key management and rate limiting for inGraphs API, an internal monitoring service written in Python Flask. Implemented rate limiting using indexed queries in MySQL, and measured different SQL queries with timing and EXPLAIN queries for tables with 200K+ rows.Autoalerts Simulator: Developed API endpoints in Golang that allow users to test alerting configurations faster by simulating alerts on historical data. Ported Python-based alert state-checking plugins to Golang.

Production Engineer Intern, Cache Infrastructure.Automated Linux Kernel Validation: Designed and implemented automated Linux kernel validation framework in Python to detect potential performance regressions in several Linux kernel versions against over 50 cluster and hardware configurations. Created a service-agnostic pipeline for launching A/B tests, and implemented a service-specific driver for Facebook's internal cache service TAO.Performance and Productivity: Improved CPU utilization by 5% on 3,000+ production boxes with CPU-bound workloads. Automation cut the entire kernel validation time by half (estimated), and allowed the team to iterate kernels faster.