Derek Cresswell

My name is Derek Cresswell. I'm a software engineer focused on building tools that help developers work faster and more reliably.

I started programming at 13 and got my first developer job in high school. I've always been curious about how software works, which led me to spend most of my free time on personal projects. Today, I work at Cerebras as a Member of Technical Staff.

I hold an Honours Bachelor of Science from the University of Toronto, where I specialized in Computer Science and minored in Mathematics. My academic focus was in operating systems, compilers, and language paradigms. I was able to participate in three research projects during my degree.

Resume

All of this information can be found on my resume. The up-to-date resume is available here.

Education

In 2024, I received my Honours Bachelor of Science from the University of Toronto. I did a specialist in Computer science and a Minor in Mathematics. For all 4 years of my degree I was a Dean's List Scholar and I graduated with a CGPA of 3.92.

My main focus was on operating systems, compilers, and programming languages. While I did a lot of low-level programming in my courses, I also spent a good deal of time reviewing the theoretical side of projects. This included the typical algorithm courses, but also extended to ideas in formal methods of programming, graph theory, parallel computing, network architecture, and machine learning.

Lucky for me, I was also able to take a number of breadth courses outside of the sciences. I had a lot of fun in a science fiction film study course as I would get to geek out with my classmates about the intricacies of film (check out my Letterboxd!). I also did two literature courses in science fiction, so I guess that I am a bit of a science fiction wiz! Another very interesting set of courses I did was on the philosophy of science. Essentially these were talking about the scientific method, why we use it, and how we can determine whether we have actual knowledge!

No matter how focused you are on your field, I think it is important to take courses outside of your main area of study. Not only does it expand your horizons, but it also lets you breathe a bit and this can be very important in managing stress.

Research Projects

During my university degree, I was fortunate enough to be able to work on a number of research projects.

Compiler-Level GPU Optimizations

My first research project was done as part of the Paramathics Research Group. This group is part of the Computer Science department at the University of Toronto and is lead by Professor Dehnavi. I received funding for this project through the University of Toronto Excellence Award (UTEA).

In this project, I worked alongside a Ph.D. student to develop new methods of optimizing matrix operations on the GPU. Our approach was to leverage the Pytorch framework to compile optimized matrix multiplications for neural networks. We found that existing neural networks had a large quantity of redundant weights. These were weights that could be removed without significantly degrading the quality of predictions. Using this, we implemented a compiler stage that would remove redundant weights until 50% of the original weights remained. This allowed for us to feed the matrix multiplications into sparse tensor cores on an NVIDIA GPU. By the nature of sparse tensor cores, this sped up the modified multiplications by 2x at no overhead cost.

Originally we determined redundant weights by zeroing our certain weights and observing changes to the predictions made by the models. Later, we began working on using formal verification to provably determine which weights could be removed. This used constraint solving methods to see if weights would get cancelled out in later calculations.

Lectures on Programming Paradigms

My second research project was done as part of a self-led research course. I was under the supervision of Professor Calver and Professor Gries.

During a course on functional programming, I was told about the various parts of functional languages. Monads, pure functions, and the like. However, at the end of the course, I asked the professor about why we would need such a language. After all, one of the points made in class was that a functional language can do everything an Object Oriented language can do (and vice versa). The answer was unsatisfactory. The professor roughly said that it let's you think differently. They obviously were unprepared for such a question. But it nagged at me. So I set out to learn more about why we have multiple programming paradigms.

My aim with this research project was to create a short series of lectures, notes, and other materials on various programming paradigms, philosophies, and techniques. These give students knowledge on the common ideologies pervasive throughout the programming world and the confidence to wield them. Since paradigms define the execution model of our programs and have a large sway over how we conceptualize our code, it should be clear that understanding these would provide a programmer with more opportunities to make informed decisions about their code. Paradigms then should be understood in terms of their strengths and weaknesses and applied where appropriate. From here, the idea leaked into broader contexts about similar programming ideas. This included Clean Architecture, test driven development, design patterns, and a few others.

After performing a survey of literature in the field, I created lectures that covered the:

• History of programming paradigms (procedural, functional, object oriented, and logical).
• How we define programming paradigms and why we use them.
• How tactfully choosing a paradigm can reduce programming effort.
• A critique of object oriented programming.
• A case study of object oriented programming in a real project.

I then presented these lectures to a class of second year computer science students at UofT as a guest speaker. This was an extremely rewarding experience as I was able to really see students connect with the material I had been working on. As a bit of a self-test, I had a few non-CS friends sit in on the lectures and asked them if they could follow along. To my delight, these friends were able to understand the flow of my reasoning and the concepts I was presenting (minus a little bit of jargon). I believe that this is one of the hallmarks of a strong presentation; that even those unfamiliar with the topic can pull something out of a presentation.

Building an Educational Rust Operating System

My third and final undergrad research project was also done as part of a self-led research course. Under the supervision of Professor Sun, I worked to build an educational operating system in Rust.

The idea of this project was to create a simple and modular OS that would be suitable for use in an OS course at university. Typically in an OS course students are stuck to writing user-mode code that simply emulates what the kernel would really be doing. While this may work for getting the idea across, something stuck out to me as missing. Letting students get their hands in an actual kernel could remove a level of abstraction from their learning but all of the popular operating systems would simply be too unwieldy. The idea to use the Rust language was made to provide a memory safe environment for students to hopefully limit frustrations that occur from memory errors.

Along with two other students, we set off to build KidneyOS. During my tenure on the project, we setup many fundamental systems of an OS kernel including memory management, paging, threading, and loading executables. My main code contributions were on the threading system where I implemented the concept of threads and the scheduler for the OS. Personally, I found the challenge of implementing thread switching to be extremely satisfying to overcome. I remember sitting in the computer lab with stacks and addresses sprawled over a whiteboard in an attempt to wrangle the various threads. This also required me to dig into how the C calling convention and the register structure of x86 processors.

By the end of this semester, we had a simple kernel that could run and switch between a few basic programs. The project was then passed onto a new group of students to continue development in the coming semesters.

Awards & Scholarships

I was grateful and honored to receive some awards during my time at UofT:

• BC Achievement Scholarship
• KSS PAC Award
• Capital News Bursary
• Woodsworth College Scholarship
• University of Toronto Excellence Award
• Dean's List Scholar (2020 - 2024)

English Language Teaching

In September 2024, I completed the Certificate in Teaching English to Speakers of Other Languages (CELTA) with the Centre of English Studies (CES) in Dublin, Ireland. I completed this course with a passing A grade, the highest grade available.

This was an intensive month-long course on both the theory and practice of teaching English as a second language. As a student, I would be taking lectures on various pedagogical approaches to learning languages and making my own classes on different parts of the English language. I would then present these classes to groups of actual students of CES who wished to practice their English. This allowed for practical teaching experience as well as immediate feedback from those learning English in how well you were able to perform your lesson. The lessons I gave and assignments I completed were done under strict requirements of quality.

My time in this course and living abroad were indescribably insightful. I was forced to teach people without being able to speak the same language. This challenge makes you rethink some fundamentals of how we learn. I truly believe that it has since had a massive impact on the way in which I communicate with my co-workers as I rely more heavily on visually communicating my coding ideas.

What Do I Program?

That's a lot of reading to get through before I talk about what code I actually write. Personally, I appreciate writing code that is for other developers, rather then end-users. Dev tools and bare-bones applications are what I do. I typically don't like to get bogged down in fancy GUIs or the latest libraries. I much prefer writing something from scratch as I find that I learn a lot more this way. You can call it being inefficient, but I call it fun.

My favorite language is C by far. While I've written C++, Rust, Python, Java, C#, JS, PHP, Haskell, and on and on, I always find that these languages feel bloated. I find the common critiques of C being unsafe or hard to use are simply ill-founded. Personally, I find a memory leak or out-of-bounds access in my code very infrequently and I attribute this to giving just a little bit extra care while writing the code. Besides, am I the only one who finds debugging kinda fun? C gives me a small, focused language that does not make a ton of assumptions or hide details for me. I find implicit code to be the enemy and revel in explicit code.

Why Do I Program?

I find mathematical proofs to be a riveting exercise for the same reason I like to program. Building and navigating these systems feels like piecing together a big puzzle. Slowly I reveal the deeper meaning and purpose of these components until in the end you have a wonderful program.

Programming, to me, is an art form. I take care with each of my brush strokes and hope to always have a masterpiece in the end. If I'm going to be working on code for the rest of my life, I might as well do it with some pride!

About the site's design?

Well, I'm a programmer, not an artist. I don't want to waste my time on some cheap tricks to make this look fancy, or reactive. I'd rather spend my time working on some interesting code (sorry web-devs).

If this puts you off from me as a dev, I'd just tell you to reflect on why that is for a moment.