Obuli Krishnaraj Venkatesan Email & Phone Number

San Jose, Ca, Us

Sunnyvale, Ca, Us

Group Identity - Designed and led the first version of privacy preserving Group Identity for Linkedin Marketing Solutions to address the privacy changes in the industry - Came up with the design that is acceptable to all stakeholders including legal and privacy - This helped to recover part of traffic that will not be accessible after the loss of Apple IDFAPost Acquisition Milestones - Key contributor in meeting the post-acquisition milestones - Designed and built a new system to improve the Drawbridge’s probabilistic Identity Graph (2X increase in coverage at 80% precision) using Linkedin’s data, Graph Algorithms and ML - Led the effort to migrate our entire custom tech stack to Linkedin’s tech stackSpark Jobs Auto-Tracking - Designed and built an outlier detection tool for engineers to collect & track 1000s of metrics from our spark jobs. This system helped catch multiple issues without any additional performance/development cost - Leveraged spark accumulators/counters that are already added by engineers to track and monitor for the outliers

San Mateo, Ca, Us

Part of the team to work on integration with linkedin and to meet acquisition goals

San Mateo, California, Us

Designed and built data pipeline to process events generated by OneSignal platform

San Mateo, Ca, Us

Drawbridge is a Cross-Device Advertising startup that probabilistically connects various devices of the same user. Early engineer that built core big data platforms for the company. Brought in Giraph and a paradigm shift to treat cross-device problem as a graph problem instead of map-reduce.Spark User Profile System- Rewrote the User profile system using Apache SparkPersonalized Cross Device Connections- My previous pairing algorithm used a fixed set of few recipes that all the IDs will go through to identify the connections.- In the newer version of the pairing each ID can chose to participate in different recipes depending on their needs.- This again is done using Giraph and it resulted in 20% - 60% increase in our pairing coverage.

San Mateo, Ca, Us

Cross Device Pairing- Generate the possible pairs for different devices(mobile/tab/desktop)- Rethought the cross-device pairing as a graph problem instead of a map-reduce problem, designed and implemented a new solution using Giraph.- This reduced the run-time from 24 hours to 5 hours with 20% increased scale and 20% increased precision.User Graph- Build clusters of devices that might belong to the same user.- Started using Apache Giraph instead of traditional MR- Started as an experiment to use giraph, now Giraph is used in multiple production systems- New algorithm brought down the runtime from >24 hours to 45 minutesUser Profile System- Designed and built a system that will accumulate all the information from various logs that belongs to a single device id over a period of time from different geographic co-location.- Used by all other platforms as baseline data.- Designed in such a manner that adding a new colo should be easy.Reporting & Conversion-Tracking- Was responsible for the revenue reporting and conversion-tracking pipelines of drawbridge.- Designed and implemented Conversion tracking system using HBase.MiscellaneousBrought in tech talk and code review culture to the company.

Sunnyvale, Ca, Us

Trend DetectionA platform to detect trending topics by analyzing the web search logs and news headlines - Designed and implemented a system that is used to detect trends for various user segments (age, gender and location) and verticals(shopping, sports etc.,) - Enhanced the quality of existing clustering algorithm - Developed a proof of concept for TimeSense algorithm that can be run on the grid using the HbaseSite ExplorerGenerate the statistics for Site owners by mining the daily web search logs - Proposed and implemented a solution using Hadoop MR to serve query level statistics - Came up with a system to determine common search queries resulting in multiple sites and compare statistics across similar sitesOozieOozie is a DAG Engine to execute MR Jobs and Pig jobs as workflows. It is an Apache open source project - Designed and implemented the logic to re-run the workflow and to skip few nodes in order to save time by re-running only failed nodes - Volunteered to be the build master for this projectMobile SearchSearch engine for mobile web pages - Designed and implemented the system to parallelize processing of crawled data - Came up with a system to use Genetic Algorithm to improve the weights associated with different document features - Designed and implemented an automated index quality checking systemMobile Web GraphUsed a Yahoo! Graph platform to generate web-graph for mobile web sites - Lead the effort to use web-graph for mobile search - Designed and implemented the framework that can be used to extract new features - Designed and implemented the workflow management to manage the graph - Came up with various features based on the links and anchor text to enrich and rank the documents

Bachelor Of Engineering (B.E.), Computer Science

Diploma, Computer Technology