Jun 2018 – Jul 2021 · 3 yrs 1 mo · Sunnyvale, California, United States

• Worked on highly scalable Hadoop-based batch pipeline for extraction and transformation of all of user activity data from multiple Verizon Media brands
• Designed and implemented a generic framework for data transformation and scalable ML model training via REST API using PySpark, JAVA, Oozie, TensorFlow, XGBoost
• Designed and developed multiple large scale data pipelines for preprocessing, cleansing and generating model training data along with data quality checks using Hadoop, Oozie, Pig, Hive, Druid
• Designed and implemented an automated classification and discovery of new referring domains used for attribution of referring traffic using Decision Trees
• Building a next generation ML-based personalization platform based on A/B testing to choose buckets based on user features

May 2017 – Aug 2017 · 3 mos · Greater Seattle Area

• AWS EC2 Systems Manager (SSM) is a service that helps configure and manage multiple EC2 instances. SSM agent is a open source software in Go-lang that runs on the EC2 instance. Project was related to the client-side diagnostics of the SSM agent. Work involved enabling the agent to stream logs generated on the instance to AWS CloudWatch, thus making it easy to access and manage the logs and diagnose issues with the agent. Systems Manager documents were created and changes were made on the agent to enable updating logging configurations of the agent right from AWS console. Additionally, secure sharing of logs with another AWS account by streaming logs to the other CloudWatch resource was completed.

Aug 2016 – Jun 2018 · 1 yr 10 mos · Los Angeles Metropolitan Area

• Worked on Domain-specific Insight Graphs (DIG) - a DARPA funded project aimed to crawl, extract information from the web, organize it in knowledge graphs using ontologies and answer questions about the domain. The project was focused initially on the human-trafficking domain data and then is being used in multiple domains.
• Introduced embedding based extractions with the precision of ~93% for extracting information from web pages
• Developed SVM based ad classification techniques with precision of ~90% for filtering useful ads
• Improved precision of geolocation extractions by 28.54% by modeling as an Integer Linear Programming problem and finding the best extractions.
• Made use of Probabilistic Soft Logic to further improve precision of extractions

Apr 2016 – Jul 2016 · 3 mos

• Worked in the Billing Products team to develop APIs as a part of next generation platform for setting up Reference Transactions and Recurring Payment profiles using Spring Framework.

Jul 2014 – Mar 2016 · 1 yr 8 mos

• Formed an indispensable asset of the team formed to analyse the transaction logs at PayPal. Work included stabilising and scaling the data transfer mechanism for pooling in more data from the data centres to Hadoop cluster for analysis, structuring the data in Hive tables to make it queriable, using Apache Pig and Hive scripts to generate error, conversion, distribution and any specific analysis and insights from the data.

Jan 2014 – Jun 2014 · 5 mos

• Worked on processing transaction logs. Work included building a stable and highly scalable data tranfer mechanism for pooling the transaction logs in the scale of TBs per hour generated in various data centres in PayPal to a hadoop cluster. Once the data is collected in the data centre, worked on making the data queriable and structured so that it can be used for analysis.

May 2013 – Jul 2013 · 2 mos · Mumbai Metropolitan Region

May 2012 – Apr 2013 · 11 mos · Pilani

May 2012 – Jul 2012 · 2 mos · Dehradun