Mar 2022 – Present · 4 yrs 1 mo · San Francisco Bay Area

• Leading the group responsible for Amplitude CDP product. Along with regular management duties, also responsible for
• Setting Vision, Roadmap and Strategy for CDP group.
• Collaborate cross functionally with finance on CDP pricing strategy, COGS, margin projections, department budget.
• Collaborating cross functionally with Product Managers, Product Marketing, Solution Consultants, Account Executives, Customer Success teams for successful sale and adoption of CDP product
• Together with team and cross functional partners, we
• Built CDP product from inception to multimillion ARR and growing…
• Built a real-time highly scalable and efficient streaming system to deliver Billions of messages to destinations in real-time.
• Built a highly scalable batch pipeline to do periodic recompute of and delivery of computed data, handling thousands of jobs per hour, delivering Billions updates throughout customer's tech stack.
• Built an ecosystem of 200+ integrations to send and receive data from.
• Successfully extended Amplitude from pure Analytics product to a platform that can be also used for * * * Behavioral Data Collection and Activation throughout the customer's tech stack.
• Built behavioral prediction layer, which leverages ML models to predict propensity of users performing an action based on past behavior.

Dec 2020 – Mar 2022 · 1 yr 3 mos · San Francisco Bay Area

• Leading the teams responsible for data collection, real-time data processing, data import and export platform and amplitude connectivity to data warehouses and partners. Working with awesome individuals to build the infrastructure to support rapid growth of Amplitude, set the direction and build the product to 10x the ease of getting data in & out of Amplitude and the incremental value customers can achieve on top of enriched data. Together with team, we
• Introduced the pattern of self-service DWH import and export. This feature is highly successful, used by most of Amplitude customers. This allowed Amplitude to grow beyond pure Analytics destination setting stage for CDP product later.
• Created an open published API and process for partner submission of bi-directional Amplitude integration. This pattern has been highly successful and contributed to 60+ integrations to Amplitude Platform.
• Introduced pattern of config based real-time json transformation to ingest data into Amplitude from any tool that supports Webhooks.

Aug 2018 – Dec 2020 · 2 yrs 4 mos · San Francisco Bay Area

• Lead the team responsible for data collection, real-time data processing, privacy/deletion and export areas. Worked with talented individuals to form the team.
• Engage PLE and ground work for integrations across marketing and ad stack (https://amplitude.com/engage)
• GDPR, CCPA compliance
• Redesign of few core pieces of data processing system
• Data import frameworks etc...

Oct 2016 – Aug 2018 · 1 yr 10 mos · San Francisco Bay Area

• Worked on back-end powering Behavioral Analytics on Trillions of Events.
• Designed and led POD responsible for Insights Product line (https://amplitude.com/insight)
• Redesigned and implemented back-end powering searchable user timelines.
• Designed and implemented Autoscaling on our largest stateful Event processing cluster, processes 2M data points per second at peak.
• Rewrote Event Ingestion servers in Java.
• Led COGS project saving ~20% of AWS cost in 6months.
• Environment: Java, Python, Cassandra, Kafka, Redis, AWS, Lambda Architecture, Distributed Query Engine, Column Store, Batch, Map Reduce, stream processing

May 2011 – Oct 2016 · 5 yrs 5 mos · Sunnyvale, CA

• Product: Riverbed SteelFusion
• Worked as lead engineer in SteelFusion Blockstore team. Responsible for technical design direction, leading feature delivery teams along with coding. Also acted as working Scrum Master.
• Riverbed SteelFusion is the first and only hyper­converged infrastructure purpose­ built for the efficient management of remote and branch office (ROBO) IT. SteelFusion combines storage delivery, server virtualization and industry­leading hybrid WAN optimization technologies into one, eliminating the need for physical servers, storage and backup infrastructure at ROBO locations.
• As part of pre 1.0 team of engineers, involved in developing and hardening log structured file system, SSD based Write Back Cache, SSD based Read Cache, High Availability and Asynchronous Replication over the WAN.
• Developed design changes to reduce RMW penalty.
• Enhanced hashing function and reduce startup time of stand alone Steel Fusion appliance.
• Proposed design changes and led Code refactoring / lock braking to improve ​read/ write performance by 3X.
• Completely redesigned failover algorithm to make failover constant time operation independent of appliance storage capacity.
• Designed and implemented POC of intelligent prefetch which works based on past access
• patterns instead of reverse engineering ondisk layout.
• Led project to implement efficient read cache simulation module to measure effect of
• increasing SSD size and policy changes on cache hit rate in production system.
• Designed Replication QoS and led feature team of four.
• Initiated scrum process an​d acting as working ​Scrum Master for Blockstore team.
• Environment:
• C++, Python, Application programming, Linux, Block FileSystem, Asynchronous
• Replication, HA, Multi Threading, Multi processor, ESXi, Distributed Systems, Key-Value store

Mar 2008 – May 2011 · 3 yrs 2 mos · san jose

• Product: Quantum DXi & EMC DL3D series of products
• Description:
• Quantum DXi / EMCDL3D is a disk based backup product with deduplication feature. It provides tape interface through VTL library and also supports backup through CIFS and NFS interface. In addition, it also has features like Path­to­Tape, OST, Replication etc...
• Environment:
• C, C++, Kernel Programming, application programming, Linux, Stackable File System,
• Backup, NFS, CIFS, VFS, Multi Threading, Multi processor, Shell Scripting.
• Roles and responsibilities:
• Responsible for enhancing and maintaining QBFS stackable file system in Linux Kernel.
• Designed and implemented I/O statistics, Ingest funneling and Space statistics features in QBFS.
• Enhanced and maintained infrastructure needed for the integration of dedup engine with
• file system to drive deduplication process.
• Worked with OEM customers and sustenance team to quickly resolve customer issues.
• Worked on maintaining and enhancing the replication functionality of the DXi product.
• Worked as a contact point for QBFS, Dedup infrastructure and Replication features. Gave
• training on features and best practices to sustenance, solution architects and other teams.
• Closely worked with other teams in defining APIs and contributed to designs.

Jul 2005 – Jan 2008 · 2 yrs 6 mos · Pune/Pimpri-Chinchwad Area

• Product: VxFS 5.0 on HP-UX 11.31 and HP-UX 11.23
• Description:
• The VERITAS File System, (or VxFS, called JFS and OJFS in HP-UX), is an extent-based cluster file system.
• Environment:
• C, Kernel Programming, HP-UX 11.31, Cluster File System, Unified File Cache, VFS, Multi Threading, Multi processor, Shell Scripting.
• Roles and responsibilities:
• Solely responsible for designing and developing VOP_SHRLOCK interface for VxFS. This allowed local and remote (NFS) processes to set share reservation on files belonging to cluster mounted VxFS, by specifying the access mode and which types of access to deny for other processes.
• Worked as part of team to integrated VxFS with the new Unified File Cache interface.
• Worked closely with HP to resolve issues quickly for the timely release of the project.
• I was part of the team, which is responsible for analyzing of feature requests. As a part of this team provided feasibility report and high level design for them. NFSv4 delegation support, cachefs, Cell OL* support, encryption support were some of them.
• Worked closely with HP to integrate VxFS and ODM with SG.
• Enhanced VxFS to work with blocking file lock requests over NFS.
• Took the ownership of the ODM product on a short notice and successfully completed the release.
• Enhanced ODM to dynamically move from single node mode to cluster mode and vice versa.
• As owner of ODM, provided assistance to other teams for all the ODM related issues.
• Worked as part of the team to increase stability of the product.
• Worked on kernel panics, hangs, data page faults and data corruptions.
• Developed automated test cases to test compatibility of VxFS5.0 with the previous releases of the product.

Jul 2004 – Jun 2005 · 11 mos

• Description:
• Current database systems lack the capability of searching across languages. This project aims at solving above problem in PostgreSQL by adding new operators to PostgreSQL database engine. These operators work across the languages by allowing users to retrieve relevant answers from Database in multiple languages with just a single query. Cost models of the operators are also added to PostgreSQL optimizer to make it powerful enough to select optimal plans.
• Environment:
• C, Perl, PostgreSQL, SQL, PL/SQL, Linux.
• Achievements:
• Implemented new Multilingual Query Operators inside PostgreSQL Database Engine.
• Published a paper in ICDE using the results obtained with this implementation.
• PUBLICATIONS
• On Pushing Multilingual Query Operators into Relational Engines, ICDE'06
• A. Kumaran, Pavan K. Chowdary and Jayant Haritsa