What's it like to study for a PhD at EPCC?

After graduating with a Computer Science degree from the University of Bath, I was unsure what to pursue next. Despite having a dissertation focused on hardware for machine learning, it was hard to find meaningful work in this area. I’ve always enjoyed learning exactly how computers work, from the compilers to the underlying architecture, and I wanted to work on something meaningful that wouldn’t abstract these details away.

Wanting to specialise in high-performance systems, I explored various programmes and discovered EPCC and its masters programmes in high performance computing. With its strong reputation, history of teaching, and focus in this area, EPCC was the obvious choice for further study. Despite having never considered a PhD before, I looked into their suggested PhD projects to get a sense of the research being undertaken. One project immediately captured my interest – researching compilers for specialised hardware. 

After a very last-minute application, I started my PhD programme, where my work has focused on compiling high-performance codes for RISC-V-based novel architectures. RISC-V is a modern, open-standard instruction set architecture that has already seen adoption in over 10 billion devices. Recently it has started being used in custom hardware accelerators, which differ dramatically from the traditional Von Neumann architectures we are familiar with today.

Machine learning accelerators

My research has been focused on a family of machine learning accelerators from Tenstorrent. These encourage the storing of data locally near each compute unit, with compute units sending data directly between each other, rather than through the slower main memory. The hope is that these accelerators will play a similar role in future HPC systems as GPUs do today, improving power efficiency and speed for particular problems. However, their vastly different underlying architectures makes programming for them difficult. My work aims to solve this issue through the compiler stack.

My first year has been filled with invaluable experiences, such as a week-long research collaboration at Tenstorrent’s Silicon Valley office with my PhD supervisor Nick Brown, where I integrated my work with their technologies. Throughout the week we had some great discussions with their incredibly talented team, and we’re hopeful that this work will lead to a publication.

Opportunities

I’m excited to be involved in a few events over the coming months: a week-long HPC summer school, hosted in Portugal; being a student volunteer at ISC 2025 in Germany; and being a support presenter for a tutorial at IEEE Cluster 2025, to be hosted by EPCC. Being at EPCC introduces an abundance of opportunities like these where I can develop skills, travel the world, and meet other students and researchers working in this field.

With nearly 20 current PhD students working in related areas, EPCC has a lively, active research community based at the University of Edinburgh. We regularly host research seminars, showcases, and clubs where we practise presenting, share feedback, and learn about each other’s research. My experience so far has been very rewarding; I’ve learnt a huge amount, had fun, and built a network of exceptional people. Studying at EPCC has been amazing so far, and I’m excited to keep growing as a researcher and see what’s next.

PhD study at EPCC

Research at EPCC

Author

Jake Davies, EPCC PhD student
jake.davies@ed.ac.uk