Hi! I'm Jacob, an M. Sc. student in compilers for machine learning at McGill University. I'm expecting to graduate in May 2023 and I'm seeking full time opportunities.
I am a member of the Compilers and Accelerator Synthesis Lab under the supervision of Professor Christophe Dubach. My research interests broadly include program transformations, static analysis, and automatic differentiation, and I've been using MLIR to do all of those things.
In addition to my research, I'm a lifelong musician. You can also find me rock climbing, weightlifting, cooking, or doing yoga.
Publications
LAGrad: Statically Optimized Differentiable Programming in MLIR
Mai Jacob Peng & Christophe Dubach
ACM SIGPLAN International Conference on Compiler Construction (CC '23)
Projects
Enzyme
I'm working on making the first hackable AutoDiff engine by extending Enzyme to MLIR. Our goal is to make it ergonomic for non-AD experts to extend the core AD engine to differentiate arbitrary dialects.
HelpMeHelp.Me
I led a team of students and alumni volunteers to build an accessible, online database of mental health resources in the Greater Toronto Area.
As part of this, I wrote a beginner's guide to JavaScript to quickly ramp-up student developers without prior web development experience.
Presentations
Hackable Autodiff: Extending Enzyme to MLIR for Reverse Mode Gradients
February 23rd, 2023
Co-presented with Martin Eppert
Enzyme Conference, Boulder, Colorado
LAGrad: Leveraging the MLIR Ecosystem for Efficient Differentiable Programming
November 15th, 2022
19th Workshop on Compiler-Driven Performance, Toronto, Ontario
LAGrad: Leveraging the MLIR Ecosystem for Efficient Differentiable Programming
November 8th, 2022
US LLVM Developers' Meeting, San Jose, California