Jonathan Ryu

Software Compiler Engineer Pey Co-op @Amazon Web Services (AWS)

Toronto, ON, CA

Enabled FLUX.dev model on “TRN1” and “TRN2” instances through use of tensor parallelism libraries, and optimized from 12 seconds to 6 seconds per image before handing to frameworks team- Delivered framework solutions for optimized partitioning of model intermediate representations- Wrote kernels using Neuron Kernel Interface (NKI), a Python-based language developed in the team- Above optimizations reduced latency by 5%, reaching 3.5 seconds per image (after frameworks team’s work)- Monitored and fixed regressions for diffusion inference models, including Stable Diffusion, using kernelized solutions

Harrison Trimble High School

99% (1st Semester, Senior Year), Top of Class all 4 years

University of Toronto

Bachelor of Applied Science - BASc