Bootstrapped a new distributed SQL database infrastructure (TiDB) from scratch at LinkedIn and scaled from 0 to millions of QPS and PBs of storage, powering critical platform use cases. Contributed to open-source to improve TiDB's extensibility and operability.
Improved LinkedIn's distributed document database's (Espresso) scalability by 10x, reduced long-tail latency by 75% by rewriting connection protocol between load balancers/routers and storages.
Made Espresso more resilient, improved availability by xx% during single load balancer/routing node failures. Built fault-injection framework leveraged by multiple teams.