Nov 2022 – Dec 2023 · 1 yr 1 mo

• Travel, personal goals and time with the family!

Jun 2021 – Nov 2022 · 1 yr 5 mos · San Francisco Bay Area

• I led the key value (Manhattan) and NewSQL (CockroachDB) charter managing a group of about 24 team members across the US and UK.
• Manhattan: Our homegrown distributed, HA key-value database provides 3 flavors - Eventually consistent that is used for most usecases, Strongly consistent and Read-only for serving batch processed data. We store petabytes of data and serve millions of QPS. Led the team to deliver on scaling Manhattan to the 3rd datacenter, privacy and data protection initiatives such as data retention, data wipeout, service to service authentication, and features for data governance, building cloud integrations for interoperability between manhattan read-only in datacenters and cloud.
• NewSQL: Bootstrapped the NewSQL charter at Twitter by providing a Twitter integrated CockroachDB as a service for internal customers and to replace Manhattan's strongly consistent offering. Built a team of 7, managed the vendor partnership and licensing.

Jan 2019 – Jun 2021 · 2 yrs 5 mos · San Francisco Bay Area

• Led a part of the Manhattan team focussed on the backend systems that power our home grown distributed key-value database - Manhattan, such as storage engines, topology management, backups and disaster recovery. Delivered on a complex storage engine migration from a home grown storage engine- MHBTree, and SSTable (Cassandra based) storage engine to RocksDB storage engine without impacting production traffic or incurring incidents, that helped us unlock more QPS per node and reduced read latencies by about 17% and write latencies by about 25% and reduced disk footprint by about 45% across all read-write Manhattan clusters. This also resulted in permanently deprecating MHBtree storage engine thus simplifying Manhattan build and read-write deploys that now uses RocksDB storage engines. Moved Manhattan backups from on-prem HDFS based backups to the cloud allowing us to reduce our hdfs usage by 60%.

Aug 2017 – Jul 2018 · 11 mos

• Managed the Cache team that offers distributed in memory key value storage solutions for Twitter. Our Redis and Memcached based solutions cater to most of Twitter’s caching needs.

Feb 2015 – Aug 2017 · 2 yrs 6 mos

• I work on Twitter's Redis and Memcached based distributed caching solution, which is among the largest of it's kind in the world (see my talk at RedisConf 2017 for more details - https://youtu.be/QznaOSk20nU)
• Some of my specific contributions include:
• Designed and implemented the data rehydration service for replicated Redis based cache clusters. This reduced the cache warm up time from many days to a few hours, significantly reducing the window of vulnerability for data loss.
• Rewrote the Twemcache cluster manager to support better health checks, cluster migrations and cluster expansion without violating data availability SLAs. This reduced the number of production issues and pages to engineers for the component by 90%.
• I own the cluster managers for Twitter’s distributed in memory caching solutions (built on Memcached and Redis), which are reponsible for all topology management.
• The Twitter infrastructure team was running on a version of Redis forked from v2.4 since 2013. I upgraded it to v3.2. This involved planning, qualifying, porting over Twitter specific changes. All of our Redis clusters have since been upgraded without any issues.
• Contributed to Redis cluster re-architecture in order to reduce annual infrastructure cost by more than $1M.

Aug 2011 – Jan 2015 · 3 yrs 5 mos

• As part of the Windows Core Storage and File Systems team, led quality efforts around integration of Storage Spaces (virtualized fault tolerant storage layer) with Windows Failover Clustering that now provides highly available storage in Cloud Platform System (CPS), the private hosted version of Azure Cloud Platform.
• Implemented software drivers to simulate storage enclosures, which reduced dependence on expensive hardware, preventing expenditure and reducing testing and shipping time.
• Worked on implementing SCSI-3 Persistent Reservations for an internal storage driver, which allows for arbitrating access to the shared disks when used in clustered systems. This helped reduce expenditure on hardware for storage spaces development and testing during Windows 8.
• Developed test plans, tools and automated tests, performed user and kernel mode debugging for various scenarios in the integration of Storage Spaces with failover clustering in Windows 8 and Windows 8.1.

Jul 2007 – Jul 2011 · 4 yrs

• Developed a framework that abstracted scenarios via UI automation and invoking the API layer. This allowed testers to construct their test automation once through simple framework library calls and based on the mode of execution (UI or API), the tests would exercise different scenarios through UI or API. This allowed the test team to focus on building scenarios , rather than spend time on the complexities of UI/API automation. This simplified testing of Windows Hyper-V Replica product for disaster recovery.
• Developed test plans, and UI automation tools and libraries for consumption by other UI testers in the team for test automation of Virtual PC in Windows 7. Led accessibility and internationalization testing efforts across the UI test team.