Yun Zhang

About Yun Zhang

A versatile software engineer specializing in large-scale distributed systems under Data Platform, Data Infra and Model Training Infra domains, and leading initiatives from conception to completion. Passionate about designing and building platform/infra services with high performance, great extendability, and strong AI envisions.Passionate about new technology, fast to learn and always be a team player.

Experience

1. Senior Software Engineer

   Linkedin

   Mar 2021 — Present · Sunnyvale, CA, US

   Established a scalable Lucene search solution in the Azure environment to build the search index and decoration data in Hadoop & Spark jobs with E2E encryptions, and deploy the apps on Azure VMs to serve production traffic. The E2E build time for a new use case has been brought down from 8+ weeks to less than 2 weeks.Architected scalable entity resolution and data enrichment workflows on Azure, integrating LinkedIn\'s on-premises systems with Azure Data Storage & Factory. Delivered a standardized solution for the data platform team to augment external CRM datasets with LinkedIn datasets, ensuring compliance with LinkedIn\'s rigorous standards and projected to generate millions of revenue per year in revenue on MSFT.Drove the K2 data platform migration from Azure to LinkedIn\'s on-premises infrastructure, achieving ~$1.1 million monthly savings and enhanced data flow efficiency. This involved transitioning Spark and Hadoop data pipelines to LinkedIn Azkaban, replacing Java indexing apps with Samza SQL streaming, and migrating online search verticals to LinkedIn\'s standard serving and federation setup, all deployed with zero downtime.Successfully developed LI Hadoop Encryption-At-Rest capabilities with the Hadoop key management proxy service on top of the LI secure infra, facilitating HDFS transparent encryption and delivering seamless, secure storage solutions for 20K+ highly sensitive production datasets across general applications.Delivered the standard incremental and online Tensorflow model training solutions via incremental offline datasets or Kafka streaming topics in production, which facilitates 5+ major recommendation model use cases with faster model releases, lower retraining overheads and lightweight onboard process.

Education

• Fudan University

  Bachelor of Engineering - BE, Computer Software Engineering

  2013 — 2017

• UCLA

  Master's degree, Computer Science

  2017 — 2018