Feb 2018 – Sep 2020 · 2 yrs 7 mos

• I led 300+ world-class engineers in AI frameworks & platform, aiming to provide highly efficient and easy-to-use AI frameworks & developer platform to revolutionize Facebook products, research and open source community. We have built and deployed Caffe2, and later PyTorch, broadly across all Facebook data centers, billions of mobile devices, and millions of AR/VR devices. PyTorch powers the entire Meta family of applications, integrity, ads, newsfeed, and Metaverse. To sustain the exponential data and compute growth of ML workload, we develop AI compilers to drive orders of magnitude performance acceleration through SOTA hardware. We are taking the ambition to supercharge the velocity of model exploration, and time to production through a powerful ecosystem and platform purpose-built for PyTorch.

Jul 2015 – Jan 2018 · 2 yrs 6 mos

• I led Facebook warehouse product infrastructure team, driving innovations in data pipeline management, metrics framework, and logging infrastructure. I also led the entire warehouse capacity planning, overall efficiency and demand control program.

Aug 2011 – Jul 2015 · 3 yrs 11 mos

• We are building the next gen data life cycle management platform, including an unified data ingestion framework, solving complex data integration problems to power data driven analytics products.

Aug 2011 – Aug 2013 · 2 yrs

• Built LinkedIn's distributed data serving platform -- Espresso

Apr 2008 – Jul 2011 · 3 yrs 3 mos

• IBM Smart Analytics Optimizer v1.0: an extremely fast data processing over compressed columnar-data in a large-scale data warehouses for BI.
• Shark: offloading computation from an RDBMS to a storage controller.

Sep 2007 – Mar 2008 · 6 mos

• Policy-based monitoring and diagnostic component for DB2 for z/OS v9.

Jan 2006 – Aug 2007 · 1 yr 7 mos

• StorageDB: embedding a light-weight DBMS into a storage controller to achieve rich funtionalities and high performance.