Aurora mysql

- Build Aerospike Graph, as part of the core Graph team. Implemented bulkloading/ write path of the graph database.- Leading Aerospike connect for Spark and Aerospike Connect for Presto/Trino products.- Collaborate with product management, quality engineering, performance engineering and documentation team and Solution architects to ensure connector is easy to use, well tested and well documented.- Developed Aerospike spark connector from scratch. List of developed features are here[https://docs.aerospike.com/docs/connect/processing/spark/index.html].- Few major features are secondary index based query processing, predicate pushdowns, native Aerospike expression support, rate limiting, streaming read and write etc..- Trino connector: https://aerospike.com/products/connect-for-presto/

•Developed product feature to support custom user defined data tagging operations across multiple components like marketplaces, websites. This system uses MongoDB for maintaining customer profile related metadata, Elasticsearch for storing tagging related information and RabbitMq for communication across services. These microservices have been developed using Spring.• Developed product feature to detect counterfeit product listings in online marketplaces. This service uses spark streaming to consume data from RabbitMq in near realtime and classify listings using spark MLlib Random forest. Spark job have been implemented in Scala.• Contributed to development of a new product High Value Target(HVT) from grounds up. This product automatically link sellers across multiple marketplaces using their key attributes like email, phone number etc, internally representing them as a graph node. Developed data ETL pipeline jobs using Chronos, spark clustering to establish link between graph nodes, Spring microservices to retrieve and aggregate data from Elasticsearch for UI consumption.• Developed and productionalize novel algorithms to extract data from unstructured datasources like E-commerce websites, search engines etc. Service exposes endpoints for parsing listing and product pages. This service is the backbone to all brand related efforts of the organization.

Researched and developed NLP system to extract and tag financial data from unstructured financial data sources. This system can massively speedup algorithmic trading. This work was awarded NSF funding of 1 million USD.This system allows to query rich financial concepts from unstructured data. Some examples being calculate ratio of Cash to Assets for software companies, what is the debt to equity ratio in this quarter? etc.

Implemented Bayesian machine learning models and proposed new statistical models for text mining and recommendation tasks.Studied various extensions of topic models and recommendation systems based on probablistic matrix factorization techniques

Contributed to full product life cycle of TransFort - an Access Control Server (ACS).