Software Development Engineer
• Migrated a complex computer vision architecture comprising 9 deep learning models to Nvidia Triton Inference Server, resulting in an average 11% reduction in latency and 13% decrease in GPU utilization per inference request.• Deployed the Triton server on AWS EC2 (Linux AMI), hosting both client and server containers to handle continuous video streams with efficient complexity management.• Applied strong knowledge of object-oriented design, data structures, and algorithm design to develop and implement an ensemble model architecture (DAG) using Triton’s Python backend, optimizing model management and scalability.• Automated testing by developing Python and Bash scripts, refactoring existing code for streaming video sessions to send inference requests to the Triton server, ensuring compatibility for production implementation and testing.• Developed a custom Docker container, resolving package conflicts and creating the Dockerfile for the Triton Inference Server.