Almog Gavra Email & Phone Number

Menlo Park, California, Us

Responsive is the best platform for managing Kafka Streams applications. Responsive solves the typical reliability, scalability, and developer productivity problems faced by Kafka Streams users by offloading the hard distributed data problems to managed services which plug into existing Kafka Streams applications. We provide the observability, management and other features every team running Kafka Streams needs, saving years of work.If that resonates, get in touch: https://calendly.com/agavra/30min

Mountain View, California, Us

During my tenure at StarTree I designed and implemented core features for Pinot's multistage query engine.- I implemented and designed the threading model for cooperative work scheduling, which fixed deadlock issues and improved query throughput for high latency queries- I improved query latencies by up to 30% for complex queries by implementing a concurrent query execution framework- I contributed over 50 commits (https://github.com/apache/pinot/pulls?page=1&q=is%3Apr+author%3Aagavra+is%3Aclosed) covering new features and critical bug fixes

Mountain View, California, Us

I led a team working to build a cutting edge streaming database for real-time applications:- Orchestrated the General Availability of database functionality for ksqlDB in Confluent Cloud- Sourced, hired and onboarded new engineers to the team- Lead the development of the team roadmap in close partnership with senior engineers and productWhile managing a team of engineers I continued to remain technical and contributed various complicated bug fixes, stability improvements and minor features.

Mountain View, California, Us

I was an early engineer in ksqlDB, a streaming database engine, and contributed across the board to the core distributed system infrastructure. During my time as an engineer, ksqlDB went from it's first public release to a production available cloud service. Some snippets into my main projects:- Built a mechanism to upgrade streaming applications logic without downtime (https://docs.ksqldb.io/en/latest/how-to-guides/update-a-running-persistent-query/)- Lead development of multiple column key support, an extension of the SQL standard to help model Kafka data- Designed and implemented multi-way joins- Expanded our serialization framework to support Protobuf- Implemented major capabilities in the User Defined Functions framework such as variadic, generic and structured type supportYou can see all of my contributions on GitHub https://github.com/confluentinc/ksql/pulls?q=is%3Apr+is%3Aclosed (at time of writing I had 357 commits, so reach out to me if you want me to point you to some interesting ones!)

Sunnyvale, Ca, Us

- Developed a zero-copy mechanism for accessing search document data, improving old-generation GC by nearly 500% and improved P99 latencies by 15%- Migrated flagship search candidate retrieval to a framework for rewriting queries that sets the foundations for moving a heuristic model to a machine learned one

Sunnyvale, Ca, Us

I was a tech lead of the Search Infrastructure team at Linkedin, and was responsible for both individual contributions and guiding junior members. - Designed, implemented and led migration effort of schema sharing across different search verticals. This effort enabled sharing of relevance features across different teams, which in conjunction with other efforts, dramatically improved key metrics.- Led the effort to migrate search update stream processing stack to LinkedIn's next-generation changelog capture, Brooklin- Architected a system to deploy terra-bytes of index data to thousands of machines without experiencing any search down-time.- Built tools to analyze query performance, improving tail P99 latencies by over 30%.- Redesigned the geolocation search algorithm, improving performance and accuracy and while reducing index size.

Sunnyvale, Ca, Us

I demonstrated end-to-end execution on projects that moved the needle on live indexing, deployment speed and relevance for LinkedIn Search. I always prioritized craftsmanship and quality design in my projects, not only delivering quality code but also refactoring surrounding code in preparation for my projects. Search Infrastructure:- Achieved 6x increase in indexing speed of new events by multithreading our indexing pipeline using the LMAX disruptor to manage thread communication- Developed facet counting algorithm that improved average search performance by 10%- Improved product deployment from 5 to 2 hours by discovering, analyzing and fixing bottlenecks Relevance & Tracking:- Implemented features that were pivotal in increasing "exploratory people search" CTR by 2-3% (searches with no specific person)- Designed and implemented a pipeline that aggregated and collected statistics on all mobile tracking events at LinkedIn (page views, clicks, etc...)

Redwood City, Ca, Us