Michael Masenheimer

Undergraduate Research Assistant @CERN

Tucson, AZ, US

Create and test convolutional and feedforeward neural networks built in C++ on AMD\'s Versal AI Core Series engines- Cross-compiling trained networks such as resnet-50 and resnet-18 through the vitis ai docker container, while working on writing my own network- Running VART testing for evaluation of FPS, accuracy and DPU usage as well as quantization validation testing for fine-tuning the precision/speed and examining data flow between PLIOs and kernels- Examine how network size impacts latency- Comparing stats of my network and resent 50 on the VEK 280 engine to field FPGA data used on the Large Hadron Collider, which examines what the future of AI engine computing could look like for CERN