Rugved Mhatre

Graduate Research Assistant

New York University•  August 2024 - Present•  New York

Developing functional and timing simulators for performance analysis of the Ring Processing Unit, a vector-architecture processor for Privacy-Preserving Machine Learning. Collaborating with PhD students to optimize Number Theoretic Transforms (NTT) and Inverse NTT on the B1K Instruction Set Architecture, enhancing processor efficiency for Full Homomorphic Encryption; research paper in progress.

Graduate Adjunct

NYU Center for K-12 STEM Education•  June 2024 - August 2024•  New York

Instructed and mentored 75 K-12 students in a curriculum focused on Machine Learning fundamentals and Deep Learning, with hands-on projects in Scikit-Learn, TensorFlow, and PyTorch. Facilitated interactive learning experiences, earning exceptional feedback from students and the NYU Center.

DevOps Engineer II

Oracle Financial Services Software•  September 2022 - June 2023•  Mumbai

Led a team in the development of a transformer model (BERT) designed to analyze user comment data and identify bugs. Developed a novel customer origination tool, resulting in a 75% speedup. Implemented a caching logic that optimized the performance of all scripts, resulting in 15-minute execution time reduction. Initiated the refactoring of Linux-based Bash scripts to OS-independent Python scripts for performance improvement in the development pipeline. Trained five recruits by conducting knowledge-sharing sessions on an overview of the codebase and the proprietary tools and technologies being used in the project. Consistently acknowledged as a top performer thrice, with recognition from peers and clients for exemplary work.

DevOps Engineer I

Oracle Financial Services Software•  October 2020 - September 2022•  Mumbai

Implemented a concurrency algorithm for the execution of test cases and stress-tested our servers with more than 200 sessions at a time, achieving an exceptional 72 hours reduction in the total testing time of 471 test cases. Streamlined execution workflow, reducing 30% waiting time by improving the queuing logic to handle execution priorities and resource interdependencies. Designed an efficient algorithm by implementing a concurrency logic to transfer files over the network, thereby improving the speed of database backups by 50%. Created scripts for database installation, configuration, and cloning, resulting in 80% fewer time delays and reducing the dependency on the database team.