Emile Courthoud is a co-founders at Nebuly and maintainer of Nebuly’s open-source projects. Emile has an engineering background acquired at ETH Zurich and work experiences in healthcare and space sectors. He loves the mountains, where he enjoys hiking in the summer and skiing or snowboarding in the winter.
Nebuly is an AI open-source company helping enterprises to build efficient AI models, slashing computing costs, reducing complexity and moving models into production faster. Our relentless focus on software to hardware low-level optimization allows us to accelerate deep-learning computational speed during both training/inference and to improve model accuracy. We have a dual presence in the US and Europe, a strong team from the world’s best universities (MIT, ETH, etc) and a fast growing community with members in 40+ countries that supports us in our mission to build the fastest AI ever.
In the talk, we will explore the challenges that developers often encounter during model deployment and strategies to address them and accelerate inference. We will discuss the performance that can be achieved with post-training optimizations and present nebullvm, Nebuly’s open-source inference accelerator.