About this opportunity:

Master Thesis: Optimal hardware selection for AI-model training

Ericsson is a world-leading provider of telecommunications equipment and services to mobile and fixed network operators. Over 1,000 networks in more than 180 countries use Ericsson equipment, and more than 40 percent of the world’s mobile traffic passes through Ericsson networks. Using innovation to empower people, business and society, Ericsson is working towards the Networked Society: a world connected in real time that will open up opportunities to create freedom, transform society and drive solutions to some of our planet’s greatest challenges.

In the Ericsson research area Artificial Intelligence (AI) we study and develop Machine Learning and AI technologies for intelligent systems. Our research spans technologies for intelligent automation, novel applications of ML/AI to differentiate Ericsson’s portfolio and the study and development of frameworks for AI in products and services. The Architectures and Frameworks group in Luleå, Sweden, work with next generation AI platforms, tools and frameworks and research on AI to further improve energy efficiency in our products.

When doing AI training, it is often difficult to predict the amount of hardware resources that training different models will need. In a heterogenous compute environment with multiple GPU types, CPU, and memory configurations, it is very easy to take a bad decision when selecting from the available hardware. Selecting the “best” hardware seems like a good choice all the time, but on a cluster-level this can leave less performant devices under-utilized, and long queues for the more sought after high-end-devices, even though many jobs would fit and be ready quicker on the lower powered devices. We would like to remove this decision step from the developer and create a support function, an AI, that can do this hardware selection for us.

What you will do:

This thesis aims to bridge the gap between mach