Oracle Cloud Infrastructure - Roving Edge Infrastructure(https://www.oracle.com/cloud/roving-edge-infrastructure/)

• Built a time-series forecasting model for efficiently managing workforce operations at eBay call centers

• Worked on building a realtime data pipeline to optimize the REST API response time from 12

seconds to 1 second using open source big data technologies - Apache - Kafka, Flink, Cassandra.

• Benchmarked and analyzed 3 databases under different loads, server capacity & use-case -

Apache Cassandra, Apache Ignite, and Elasticsearch for write and read throughput.

Core member of the product team for "Inferlytics" - It is a first of its kind integrated eCommerce search & browse experience engine which combines power of natural language processing and domain-specific knowledge bases to offer a differentiated experience for retail customers

• Worked on developing and designing the Real-time Analytics platform end to end using Big Data tools (Apache -Storm, Kafka, Zookeeper) for capturing and processing messages, detecting actionable events in the message stream, and reporting meaningful insights in the data

• Installed and configured multi-nodes fully-distributed Apache Storm, Apache Kafka, Apache Zookeeper, Apache Cassandra, Druid and Elastic Search cluster.

• Cluster capacity planning, performance tuning, cluster monitoring, troubleshooting.

• Load balancing clusters to add and remove nodes based on the traffic.

• Scripts for automating all the deployment.

Gimlet (OCR platform)

• Designed and developed a Testing tool for the testing team to validate the results from the OCR extraction. It was a REST API and Swagger UI based application. The application calculated the accuracy of the data extracted from all the receipts feed to the OCR platform and sent the detailed report by email, which was built using JavaMail API.

• Worked on creating REST API using embedded jetty in Java to extract locale from the invoices and expenses receipt which was feed to the OCR platform, the locale contained the city, country code, language and the locale value. It used Ngram approach and fuzzy matching to find the city from the extracted text.