Suhas Devendrakeerti Sangolli

• Worked on designing and implementing the deep learning-based entity match classifier that classifies continuous streaming entity pairs • An iterative method is designed to update the model parameters for the new dataset • Adopted active learning to obtain the true labels for the entity pairs • The entity pairs prediction and iterative model training are decoupled and work parallelly, independent of each other • Technology stack used: Python, PyTorch, BERT model, Kafka, Flask, MongoDB, GitHub

-Involved in the design and implementation of a high-performance ETL pipeline for the continuous supply of onboard data -Used spark transformation techniques such as aggregation, ordering and smoothing -Worked on the development of visualization charts for validating the raw and refined data -Worked on documentation (Confluence) for user requirements -Worked in agile, cross-functional research and development team -Technology stack used: PySpark, Python, SQL, Azure databricks, Data Factory, Data Lake, Git

-Worked on Business Intelligence (BI) tool called Datastage -Involved in the development of data pipeline for EOD/Weekly/Monthly ETL processes -Contributed to execution and maintenance of automated jobs to perform ETL tasks -Developed spark pipeline for onboard data refinement -Tested and reviewed the enhancement codes -Worked with stakeholders to translate the business requirements into queries and to represent the analytical reports -Technology stack used: Datastage, Servicenow, SQL, Spark, Python, GitHub

I learnt courses related to Data Science and Engineering like Data Warehouse, Data Mining,  Hadoop, Spark, ML, DL and MS Azure. Also, I have exposure to Information visualization and visual analytics. My master thesis involves the design and implementation of a DL-based classifier to classify stream entity pairs. It adopts a BERT-based classifier to classify the streaming entity pairs. It also uses other technology stacks such as PyTorch, MongoDB, Apache Kafka, and Flask.