Arizona State University Tempe, AZ, US

Foundations of Algorithm, Data Processing at Scale, Deep Learning, Statistical Machine Learning, Data Mining, Artificial Intelligence

Vellore Institute of Technology Chennai, TN, India

Data Structures and Algorithms, Object-Oriented Programming, Database and Management Systems, Large Scale Data Systems, Machine Learning, Natural Language Processing, Operating Systems, Applied Linear Algebra

Amazon Seattle, WA, US

• Proposed and led the development of an Interactive Data Extraction CI/CD Pipeline for Amazon Science, reducing waiting and processing time by 15% compared to traditional methods.

• Leveraged a suite of AWS services including Fargate, EC2, and S3 to orchestrate a robust & scalable infrastructure.

• Went above and beyond by developing a Python module to streamline access within AWS Sagemaker, enhancing accessibility and usability for the Amazon Science team.

Arizona State University Tempe, AZ, US

• Created a Constrained Masked Language Model (MLM) VarBERT, revolutionizing the recovery of meaningful variable names from decompiled source code, achieving remarkable accuracy of 50.70%, surpassing state-of-the-art models. IEEE S&P

• Pioneered methods to boost the cognitive prowess of LLMs like T5 with instruction learning. Achieved comparable performance using just 10% of the original dataset, marking a breakthrough in natural language understanding. arXiv

• Demonstrated the limitations of large language models (LLM) like GPT3 and T5 by creating numerical feasibility data set. Created data set and experimentation to show that the GPT3 has accuracy of just 19%. Also, demonstrated the knowledge ingestion inability of such models. EACL 2023

Wipro Ltd. Pune, MH, India

• Spearheaded the development of an ML-driven customer service routing system, optimizing the incorrect routing by 18% through the implementation of a hybrid model combining decision tree and neural network algorithms.

• Achieved a staggering 40% reduction in processing time for transforming millions of rows in Mastercard’s secure customer transaction big data (TBs) by engineering a robust ETL pipeline leveraging Apache Spark, SQL, and Apache Hadoop.

• Developed batch manager jobs using Unix to schedule and automate Apache Spark and SQL scripts, reducing manual labor by 24 hours per month, and streamlining data processing and enhancing efficiency.

• Mentored engineers, providing comprehensive guidance in Apache Spark and Python. Established industry best practices, directed career development, and conducted customized technical learning sessions for individual engineers. Additionally, facilitated the seamless integration of new team members into the workplace environment.

Center for Development of Advanced Computing (CDAC) Pune, MH, India

• Introduced and programmed a NER (Named Entity Recognition) model (Bi-LSTM) to route the user to the most relevant customer service professional. The model was leveraged to extract relevant features from the conversation. This implementation reduced the incorrect customer routing by 13%.