Oct 2019 – May 2022 · 2 yrs 7 mos

• I lead the team that builds Aurora database engines (MySQL/PostgreSQL) and distributed storage for the fastest growing service in AWS history. Our database aware custom distributed storage system also powers Amazon Neptune graph database and Amazon DocumentDB database services.
• We work on enterprise class features for open source databases such as: multi-master, global databases for read/write scalability, push down query execution, multiple reader instances to shared storage, server-less database offering etc. We accomplish this by running our database engines on top of a custom built distributed storage that provides multi AZ availability, scale out performance, continuous backup, parallel restores, x-region replication/snapshot copies, fast cloning, fast rewind of databases etc.
• Read/listen about our innovations here:
• https://awsmedia.awsstatic-china.com/blog/2017/aurora-design-considerations-paper.pdf
• https://www.allthingsdistributed.com/files/p1041-verbitski.pdf
• https://aws.amazon.com/rds/aurora/ and https://www.youtube.com/results?search_query=aws+aurora+reinvent

Dec 2014 – Oct 2019 · 4 yrs 10 mos

• I lead the team that builds distributed storage for the fastest growing service in AWS history: Aurora MySQL and Aurora PostgreSQL databases. This database centric storage system also powers Amazon Neptune Graph DB and Amazon DocumentDB.
• We work on enterprise class features such as: multi AZ availability, scale out performance (5x compared to other commercial engines), continuous backup, parallel restores, x-region snapshot copies, fast cloning, fast rewind of databases etc.
• Our storage is compute capable and does parallel query evaluation on the storage fleet, provides opportunistic hierarchical ledger based Multi master databases, fast physical replication across regions and global databases.
• Read about our innovative storage system here: https://www.allthingsdistributed.com/files/p1041-verbitski.pdf

Sep 2013 – Dec 2014 · 1 yr 3 mos · Seattle

• Setup a brand new team with 15+ members (s/w dev, statistics & ML) responsible for forecasting customer purchase behavior. Architected the backend data processing and services that were integrated to various Amazon retail systems to enable effective message prioritization, marketing spend estimation, and planning.

Dec 2010 – Sep 2013 · 2 yrs 9 mos

• Led a team of 15+ developers responsible for Bing local business search for multiple locales, Bing local professionals (doctor/lawyers etc) search, and a machine Learning based entity conflation platform.

Apr 2010 – Dec 2010 · 8 mos

• Designed and Implemented publishing pipeline for Bing/Shopping vertical using map reduce infrastructure. This improved the scalability (10x more documents), and performance.
• Designed and led the development of image processing pipeline for Bing/Shopping vertical using map reduce infrastructure.

Dec 2008 – Mar 2010 · 1 yr 3 mos

• Co-inventor of a scalable lookup service. Implemented this invention for Single Instancing in storage.
• Co-inventor of a Distributed Disk based recovery solution for distributed services. Implemented this invention for MSN messenger service.

Sep 2005 – Nov 2008 · 3 yrs 2 mos

• Summary:
• Worked in a startup type of environment that involved quick prototyping and adapting to changing requirements and successfully transitioned from the prototype mode to production.
• Established and led the team in adopting best engineering practices: unit testing, coding practices etc.
• Architected and implemented major portions of the product related to blob store (in entirety: persistence/replication etc), network libraries, security etc.

Nov 2002 – Aug 2005 · 2 yrs 9 mos

• Worked on scalability/security/clustering features for Distributed File System (DFS) in Windows Server 2003. Components spanned kernel drivers, and user mode services. The features that I worked on enabled DFS to become a key driver for the sales of windows for file server workloads.
• Developed windows explorer shell extension for "Previous Versions" (snapshots), and backups of a volume on Windows Vista. This feature is equivalent to iMac’s popular “Time-machine” feature.

Aug 2001 – Aug 2002 · 1 yr

• Worked under Dr.Richard Newman and worked on Steganalysis of JPEG. It involved embedding data in jpeg images to evade steganalysis and eventual discovery of embedded data.

Aug 2000 – Aug 2002 · 2 yrs

• Worked as a head teaching assistant (supervising 6 other teaching assistants) for Applied Discrete Structures course. Responsibilities include mentoring students, preparing, conducting and grading tests.

May 2001 – Aug 2001 · 3 mos

• Multicast protocol development for edge aggregation routers.
• Implemented DVMRP MIB
• Designed and implemented the major modules of IGMPv3