• Enabled Spark NLP library to optimize and run deep-learning inference using OpenVINO Runtime and showcased benchmarks demonstrating up to 40% improvement in inference speed over Tensorflow.

• Wrote Java Native Interface (JNI) bindings to expose the OV Runtime C++ API in Java and enabled support for several models including BERT, T5 to seamlessly leverage the integration through Scala and Python APIs.

• Developed nlp-benchspark: an extensible tool for benchmarking NLP inference with PySpark.

• Published technical blog posts in collaboration with Intel.

• My project was one of the ~40 selected for the Contributor Lightning Talk Series among ~172 projects.

1. Mentored a team of 3 in collaborative development of exporter services in Java for periodically ingesting performance metrics from Application Monitoring platforms like CloudWatch in order to predictively model application health.

2. Achieved up to 30% improvement in time-series data collection efficiency by implementing multithreading and caching mechanisms and cut cloud costs up to 20% by moving automation services to Serverless.

3. Built several core product features including system-wide notifications service, autonomous inference of cloud topology and customer onboarding web-app.

4. Co-developed and patented autonomous release management system to analyze production release quality.