Arm Workshop

A Practical Introduction to Optimizing AI on Arm CPUs and NPUs

16 October 2025
Time 10:50 AM to 12:50PM
Madrid, Spain
Galeria Room, Hotel Meliá Castilla

Presenters: Kieran Hejmadi (Arm), Matt Cossins (Arm)

This session will equip participants with practical strategies and tools for deploying artificial intelligence across Arm-based computer architectures. Learners will understand the ML use cases for CPU, GPU, and Arm Ethos NPUs, identifying real-world use cases for each—from control logic and vision processing to power-efficient edge inference.

Participants will gain practical understanding on using the Vela compiler to optimise TensorFlow Lite models for Arm’s Ethos-U NPUs, and learn how to evaluate deployment trade-offs across different hardware targets. The session will also introduce TOSA, a standardised operator set that supports model portability across platforms.

By the end of the session, attendees will understand how to select the right compute architecture for their AI application, and how to deploy models effectively under real-world hardware and power constraints.

Who Should Attend

Faculty and instructors in computer science, electrical and computer engineering, and related STEM disciplines interested in integrating AI and ML into hands-on teaching.

Special Equipment

While no special equipment will be required, participants are encouraged to bring their own laptops to the session.

About Kieran Hejmadi:

Kieran holds MEng in Electronic Engineering with Nanotechnology from the University of Southampton. Kieran’s career began at Sony, where he worked as an engineer in the image processing. Kieran then transitioned to the life sciences sector, joining the startup unicorn Oxford Nanopore. There, he was responsible for delivering software and hardware features for DNA sequencing spanning embedded IoT devices to server-class hardware. After 7 years, Kieran returned to the semiconductor field as a Software And Academic Ecosystem Manager at Arm, collaborating with universities and industry worldwide to bolster the application software development ecosystem used within academia.

About Matt Cossins:

Matt is an Academic Ecosystem Manager at Arm, working to plug the industry-academia skills gap, and previously held engineering roles at cellXica and Capgemini, where he worked on embedded systems hardware and software – including hardware acceleration of 5G. He has a strong interest in AI/ML and for his master’s thesis, designed and implemented a neuromorphic, chaos-based AI kernel.