Tales from the Trenches of Resource Allocation
Speaker

Meet Ekin, a passionate data scientist, and GPU enthusiast. During her Master’s studies, Ekin dove into the privacy implications of federated learning with deep neural networks while also gaining practical experience as a data scientist. Through her journey, she observed inefficiencies in GPU usage both in the research and industry settings. As a Data Scientist at Run:ai she thus established the “AI Infrastructure Club” Discord server - to discuss all things GPU-related, exchange ideas, and support each other in making the most out of GPUs. To further streamline GPU usage for data scientists, Ekin and the Run:ai team actively work on “Genv,” a fully open-source project to use and share GPUs efficiently.
SUMMARY
In the realm of AI, GPUs play a crucial role in data science projects. Yet, squeezing the most out of them is a challenge, particularly for data scientists with limited resources and time. This talk explores the often-overlooked world of GPU orchestration and allocation with a dash of humour. We’ll address common mistakes like underutilization, static assignments, and resource-sharing pitfalls. Introducing Genv, an open-source tool, we’ll transform these challenges into opportunities for seamless and efficient AI workflows.

