2021.4 - now Resource allowance systems in compute infra (infrastructure)
• Improved server utilization via building stackable reservations, which enables resource sharing across reservations.
• Drove the design and implementation of prototypes and iterations. Released the feature to the whole company eventually.
• Extended stackable reservations into GPU based workloads and L-shaped workloads. Drove tooling development for migration.
• Patent based on the work has been filed
2019.9 - 2021.4 Real-time data ingestion and analytical platform (infrastructure)
• Solved the table retention problem by designing and implementing virtual retention feature, which leveraged data warehouse to make tradeoffs between speed and expensiveness.
• Optimized query performance via raw data scanning, operators push down, full query cache, make compute and storage close, parallel partition scanning.
• Released the feature to the whole company, including 30+ table views, and 2X faster than presto based implementations.
• Patent based on the work has been filed.
• Published in VLDB 2022 (https://www.vldb.org/pvldb/vol15/p3522-mo.pdf)
2017.6 - 2019.9 Data intensive applications (infrastructure + backend + product)
• Built network insights API from scratch, onboarded use cases including cell alerts, regional maps, local maps, region selection, cell uploads, report building, etc. Used by CXO level executives in partner companies (e.g., T-mobile, AT&T, MTN) to make decisions to improve network infrastructure ecosystems.
• Ease the management of network and market metrics by designing and implementing metrics system.
• Built connectivity measurement product and backend from scratch to measure the impact of infrastructure change.
• Optimized and maintained market insights API.
Highlighted posts:
https://engineering.fb.com/2021/06/28/connectivity/digital-divide/
https://engineering.fb.com/2018/08/10/connectivity/announcing-tools-to-help-partners-improve-connectivity/
