this /static/media/twitter/CHTY3U.png

Efficient and easy data loading still remains a challenge for large-scale deep learning. As a team of DL practitioners and researchers, we've experienced many of these issues first-hand. In this talk, we'll dig into some of our pain points, share learnings we had along the way and explain why, in the end, we decided to build our own solution to the issues we faced. Our open-source library enables ML teams to share, load, and transform data in a collaborative, flexible, and efficient way. We explain how data and GPU stall can be avoided, while reducing the costs and maintaining the flexibility for deep learning research.

Dr. Thomas Wollmann

Affiliation: Merantix Momentum

Thomas is VP of Machine Learning Engineering at Merantix Labs. He holds a PhD (Dr. rer. nat.) in Computer Science and a MSc in medical computer science from Heidelberg University. He worked in various fields in academia and industry. In particular, his expertise is in the area of machine learning, computer vision, usability engineering, and engineering leadership.

visit the speaker at: Homepage