During my internship at GEICO on the Mobile Cloud Integration Tier team, I improved the Mobile Cloud API by migrating endpoints to serverless functions on Azure using C#. I significantly reduced Redis cache latency by 300% and implemented secure callback functionality. I also designed and built an automated diagnostic tool using logistic and linear regression, which improved anomaly detection in production server performance by 15%. Utilized Splunk, Python, and SciKitLearn to manage and analyze API calls on non-production servers.

Interned at Capital One's Center for Machine Learning with the Document Intelligence and Vision Engineering team. Built and trained machine learning models using Paddle for object detection on ID documents. Created a Docker image for the team's Paddle model training, published to Artifactory. Developed an end-to-end pipeline using PP-YOLOv2 and PP-YOLOE object detection algorithms achieving 98% accuracy. Utilized Python, Paddle, Docker, AWS S3, and Artifactory.

As an intern under Dr. Zhang at the University of Maryland, I addressed efficiency issues in machine learning computations. Developed a machine learning architecture that utilized data parallelism for efficient distributed processing and documented it in a comprehensive research paper. The implementation resulted in a 67% increase in efficiency when training various ML models across four nodes. Worked with Horovod, TensorFlow, AWS DLAMI EC2, and OpenMPI.

As a high school student, recruited from the competitive Open Water talent pool to work for a startup. Developed the back-end REST API for the Medius web application using NodeJS, ExpressJS, and MongoDB Atlas. Deployed and managed the MVP using AWS Route53, EC2, and VPC along with GoDaddy. Enhanced the web application by implementing OAuth integration, ElasticSearch, and filters.