Overview

Google, IBM & Meta Certificates – 40% Off

One Coursera Plus subscription covers most Professional Certificates on Coursera.

By studying this course, you will gain an understanding of the technological trends driving AI to the edge, key concepts in mobile AI including on-device processing and real-time inference, and how to train language models with domain-specific data. You will also develop practical knowledge of edge AI deployment fundamentals, including quantisation and model compression, build applications using large language models (LLMs), and gain an appreciation of the security, privacy, and ethical challenges involved. This course explores three fundamental questions shaping the future of mobile AI: what is driving AI to the edge, what tools and techniques are needed to deploy LLMs on Arm-powered mobile devices, and why Arm is uniquely positioned to harness the advantages of edge AI. Across six modules, you will move from foundational concepts in training and inference through to advanced optimisation techniques and security considerations, equipping you with the knowledge and skills to deploy and develop AI applications at the edge. The course is taught by Michele Magno, Head of the Center for Project-Based Learning at ETH Zürich and a leading researcher in Tiny Machine Learning and energy-efficient IoT, alongside Pietro Bonazzi, a PhD Candidate in Efficient AI Computing at ETH Zürich specialising in memory-efficient, low-latency systems for edge platforms. A basic understanding of programming and machine learning is recommended.

Syllabus

Module 1: Why Deploy LLMs on Arm-Based Mobile Devices?
- This first module introduces how AI is moving from the cloud to the edge—enabling faster, more private, and more efficient computing on devices such as phones and sensors.
Module 2: Introduction to Training and Inference
- In this module, you will continue to hear from Gian Marco Iodice as he discusses the evolving definitions of large and small language models in the context of today’s edge devices.
Module 3: Optimising LLMs for Deployment on Edge Devices
- This module begins with an introduction to fundamental compression methods, followed by a detailed examination of quantisation and its role in reducing model size and improving efficiency. You will also delve into pruning and other advanced techniques that streamline LLMs while preserving their effectiveness.
Module 4: Optimising Large Language Models for Mobile Devices: Performance, Memory and Power Efficiency
- This module explores how LLMs can be optimised to run efficiently on mobile devices. It examines the key constraints of mobile deployment, including compute-bound and memory-bound processing, which are critical to understanding performance limitations.
Module 5: Advanced Edge AI: Efficiency and Scalability Techniques
- This module explores the evolution of machine learning models on edge devices, tracing the progression from early architectures like AlexNet to today’s advanced transformer models.
Module 6: Securing AI at the Edge: Privacy, Ethics and Challenges
- This module explores the security, privacy, and ethical considerations involved in deploying large language models (LLMs) on mobile devices.