About
Awright!? I’m Perry Gibson, working at the intersection of machine learning📉 and systems🔩.
I recently completed a PhD at the gicLAB (University of Glasgow), supervised by Dr José Cano Reyes.
I’m interested in across-stack optimisation of neural networks, especially with a compiler-centric perspective. For example, Apache TVM, MLIR, and tensor compilers in general. In fact, that’s what my PhD was all about!
You can find some of my research work under /Publications. And other publications and posts under /Posts, and open source contributions under /oss.
I’ve undertaken research visits at the Barcelona Super Computing Centre (2019, through the PRACE Summer of HPC program) investigating optimized data placement in heterogeneous memory systems; and Northeastern University (2022, Boston, USA), exploring MLIR and homomorphic encryption.
I’m at home in the shell, and manage a suite of self- and cloud-hosted services. I like my workflows to be reproducible: e.g., through containerisation, or template systems (like cookiecutter or AI Assets). I try to learn from and educate my team members, and leave a trail of documentation like footprints.
You can read more about my work on my LinkedIn, GitHub, Google Scholar, or blog.
Feel free to get in touch with me if you want to discuss my work, opportunities, or such.
