NASA HACKATHORN PARTICIPANTION

STC participated in the NASA GPU Hackathorn based on our research in the past two years on our GPU computing testbed funded by NASA CISTO as a membership. Our team achieved remarkable milestones in the realm of machine learning optimization. Using NVIDIA’s Nsight System as a diagnostic tool, we successfully identified and addressed critical performance bottlenecks in the existing algorithms. By harnessing the computational power of a single GPU, we were able to drastically improve the code’s execution speed to be six times faster. Additionally, we are in the advanced stages of implementing innovative data loading techniques and leveraging multi-GPU within TensorFlow to further boost performance. These combined efforts have not only significantly enhanced the efficiency of the model but have also provided a strong foundation for the scientific research to support AQ studies. We hope to continue leverage the NASA GPU Hackathorn in the following months.

5 STC students/faculty and Dr. Duffy make up Team AQ for the NASA Hackathon