Mridul M. Email & Phone Number

Sunnyvale, Ca, Us

Tech lead for Spark team and Spark understanding efforts.

Wilmington, Delaware, Us

Contributions to spark started while it was still at amplabs - it was a byproduct of our work on BigML at Yahoo labs : spark was the platform we narrowed down for work.Extensive modifications to spark to get it running on hadoop - I added yarn support, multi-tennancy support to spark, scheduler enhancements, making spark process and rack aware .. among other things.While in labs, experimented with spark in house - a) Modification to leverage gpu's for vertical scalability.b) Modification to remove critical performance and functional bottlenecks in spark : for example, 2 gb block limitation in spark.c) Reduce memory footprint of spark, reduce the heavy VM memory usage, etc.

Mountain View, Ca, Us

Santa Clara, California, Us

Part of the Apache Spark team.Developed and lead Safari - a performance, monitoring and debugging tool for Apache Spark.

Sunnyvale, Ca, Us

Flickr Tripod team.- Pixel service.

Sunnyvale, Ca, Us

Mobile search team.- WebmapPrimarily focused on speeding it up by an order of magnitude.Work presented at Yahoo Techpulse'15 under Systems track.

Sunnyvale, Ca, Us

(Roughly chronologically)- YIMADesigned the system and developed backend for Yahoo! in media advertising platform.This ran a successful pilot on Yahoo! Image search.- Elections 2012Designed and developed the data mining platform for elections 2012 effort, primarily used in the signal blog. The data insights are based out of what is exposed from this system.The grid system directly generates or acquires data from a variety of sources including low latency search log mining, twitter sentiment analysis, etc.http://news.yahoo.com/signal- JMS provider for HedwigDesigned and developed the JMS provider for Hedwig; which has been contributed back to apache.https://cwiki.apache.org/confluence/display/BOOKKEEPER/HedWig- Big MLPlatform for developing ML algorithms at scale using spark http://spark-project.org/Developed state of the art distributed matrix library for both dense and sparse matrices .. supporting common matrix operations (add, scale, multiply, etc), QR factorization.Other projects under this umbrella include work on performance enhancements to inhouse GBDT (improved by roughly 18x from baseline per iteration - and 10x end to end).Implemented large scale collaborative filtering (hundreds of million rows/columns), joint non-negative matrix factorization (the matrix multiplication version of https://github.com/amantrac/JNMF)- Groups SummarizationWrote the reply parsing module in Summarization, which splits an incoming mail into various reply depth based snippets : much more precise and robust than existing yahoo groups and yahoo mail solution.Does annotation transfer from rendered view to character level over the input html/text mail.

Sunnyvale, Ca, Us

(Roughly chronologically)- PSOX projects.Designed, developed the extraction management system of psox. Prototyped a grid based version. Designed and implemented the initial version of aggregator mining and original home page (homer) projects.- Mediafaces.Designed, developed the mediafaces backend which powers faceted experience for image and web search. Designed the next gen grid based version which is currently in production.- Predictalot.Designed, developed the backend for the largest combinatorial market simulated - a prediction game for ncaa'09 (march madness 2010), fifa'10 and ncaa'10.http://predictalot.yahoo.com- PredictopusModified the predictalot backend for cricket world cup'11 for Yahoo! Cricket.Yahoo! Cricket team built the frontend on top of this to expose it as Predictopus.http://cricket.yahoo.com/cricket-games/predictopus/- IbisImplemented to work on hadoop, scale-out to grid scale, built a client-server impl based on the initial prototype of Ibis - a provenance manager for multi layer systems.- YIMADesigned the system and developed backend for Yahoo! in media advertising platform.This ran a successful pilot on Yahoo! Image search.In addition, I developed (at times with other scientists) a bunch of util's and frameworks for various tasks within research and outside - annotation api, indexes on hadoop, robust tidy, etc.Awards:Yahoo! India U Rock quarterly award for ownership April 2010

Palo Alto, Ca, Us

Primarily focused on the runtime and server side development.(Roughly chronologically)Aug 2001 - Dec 2003 :Ownership and development of the server side infrastructure and rproxy for "Sun ONE Portal Server, Secure Remote Access"Jan 2004 - April 2005Worked on teams responsible for designing the runtime infrastructure for applications to be developed with the next release of "Sun Java Studio Enterprise IDE".April 2005 - Oct 2007:Responsible for ownership and development of the XMPP based Instant Messaging server , http binding to XMPP and client side api for Sun Java System Instant Messaging.Awards:Most valuable engineer in tools group 2005.Award for technical excellence at Sun India Y'07.