Fall '20 @ Shopify

Fall '20 @ Shopify

It has been 9 months since the beginning of the quarantine. And to be frank, nothing too exciting have happened during this co-op term (which is a good thing, especially during these tough times, boredom is a privilege). But as per tradition, I will give a review about my internship experience followed by my experience on taking CO 367 (Nonlinear Optimizations). Finally, I'll end with some thoughts on our world.

Storefront Renderer Team at Shopify

Although this is my second time here at Shopify, what I worked on this time is completely different than my first work term here. As a quick reminding, my first time at Shopify I was on the ML Platform team, where I worked on creating an robust production ready platform that carries data scientists through the entire model life cycle from development to deployment. This time, however, I was on the Storefront Renderer team under production engineering sector. What I got to work on here had a lot of highly measurable and high direct impact to end users.

So, what is Storefront Renderer (SFR)? Well in short, it is a specialized application that is solely responsible for all read-only requests made externally to Shopify. So every time you visit any Shopify store, that is being served by SFR. As you can imagine, this is an extremely high stake project. The vibe I got here is mission critical, every PR is being carefully reviewed and nitpicked to make sure it won't impact performance.

I was fortunate to have the chance to work on many highly impactful functions. And by impactful I really mean IMPACTFUL. For example, one feature I worked on effects about ~20% of all storefront traffic... So, 20% of bandwidth for read-only requests across all of Shopify was impacted by my changes. That feeling when you just go on a random Shopify store and be able to see your changes right there, sure is something else!

Although, it was not all rainbows and sunshine. Since this term is from Sept. to Dec. it perfectly overlaps with BFCM (Black Friday & Cyber Monday). It is arguably the greatest event every year for Shopify. With an immense amount of stress put on to the system. And since it is 2020, and more and more people rely on online shopping, Shopify's service basically became an essential service for businesses to remain in business. So, there was a 2 months long code freeze imposed on mission critical applications. And of course, you guessed it. My team is one of them. Starting early November 2020, no code that implements new features can be added until early January 2021. Only bug fixes can be shipped with no exceptions. As you can imagine, this was not super great for interns like me. Since I had bunch of new changes made ready to ship, only to find out, they cannot be shipped until after the end of my internship. A bit of a bummer, but I understand where they are coming from.

All in all, it was a refreshing experience working on such a critical service for such a gigantic tech company. Overall, I enjoyed my time here and like always, people here at Shopify are one of the nicest group of people I've ever worked with.

CO 367 - Nonlinear Optimizations Review

I think this is definitely the most advanced and specialized math class I have ever taken. It is essentially a combination of all the math topics I have learned (with the exception of combinatorics and graph theory). So more often than not, I had to use both my calculus skills and linear algebra skills together. In conjunction with the optimizations knowledge learned in CO 250. Just when I thought CO 250's proofs are long, this course really showed whose boss.

As for what topics we learned, the name of the course is quite relevant. The course taught us many methods of constructing/solving unconstrained/constrained nonlinear programs using methods like gradient descent, trust region method, KKT conditions and etc. Really studied the fundamental theories and ideas behind machine learning methods. We even had a couple programming assignments where we implement various descent methods and curve fitting algorithms from scratch (well maybe used numpy here and there). The last chapter talked about how all of these optimization methods relates to the way neural networks are built. Which I found was pretty interesting since most neural networks crash courses online doesn't make these theoretical connections.

Anyways, this course was definite not an easy course by any means. But I think the content is extremely interesting and definite worth learning to establish a good theoretical foundation for ML.

State of Our World

2020 sure have been a tough year, probably the toughest year in the past decade. The COVID-19 pandemic have impacted quite literally everyone's life everywhere on Earth. Personally, I've been basically quarantining since mid-March. Honestly I thought the pandemic was under control in late August when cases in Canada is about ~300/day, who would've expected that now we're at ~6k/day and growing. And somehow, people are less careful than they were in comparison to April (when the cases were ~1.8k/day, aka the first wave). I think people at this point are numb to the numbers, and all just want to forget about it. However, the truth this, the virus won't go away if we just turn a blind eye. The more we chose to ignore our problems, the better it will flourish.

(START OF RANT) What I really don't understand is why there are still people out there who refuse to wear masks and stand 6ft away. Somehow people think that by telling them to wear masks for the better health of our society is evading their "freedom" or their "rights". Yes, you may be living in a free world and you should be grateful for that, but wearing masks to protect others is just the morally correct thing to do. If you think logically, what does the government have to gain from advising you to wear masks besides coping the virus and increasing the public health status and protecting the vulnerable. Had we practiced proper social distance and mask wearing in the early days of the pandemic (around February 2020), we could have already back to our normal way of life. Just look at places like Taiwan... They barely have any cases now, just because in the beginning they took serious care of the situation before it got out of hand (did I mention, their peak per day cases is 27 cases/day, yea that's right, about ~300x less than Canada). I heard there are night markets, concerts, and things are basically back to normal for them. And consider this, Canada is about 27,651% bigger in terms of area than Taiwan, yet only have 14million more people than them. We have much lesser population density, if they can do it, why couldn't we also? I know it is no use talking about the past, but my point is, the earlier we start taking serious care of this situation, the lesser people are going to get hurt/pass away, less job losses, less economic damage, less precious time wasted, and earlier we can to get back to normalcy. (END OF RANT)

On the bright side, there are bunch of vaccines are through regulatory approvals, namely the Pfizer and Moderna vaccines. I hope the government will be able to distribute them in a timely and effective manner. I saw some news that said the aim is by September 2021, whoever wants the vaccine in Canada would have a chance to get it. I sincerely hope that's true and by late 2021, we'll be able to safely go out to hang out with friends, and back to in-person school and work.

Closing Note

Alas, what a year. No one could've expected what 2020 turned out to be. Looking back at my 2A school term post, so many plans I had went south. Who knew, we were going to be stuck inside for most of the year. If I learned one thing from 2020, it is that we must cherish the seemingly mundane and "normal" things we do and take granted for every day. If you tell me in 2019 that in 2020 that I won't be able to meet with friends regularly, can't go to dine-in restaurants freely, can't travel to places as I wish. I'd think you were just messing with me. Gratefully, 2020 to me, seemed like it went by in a flash, since I'm fortunate enough to be able to safely work/study from home. Whereas, there are millions of people who still have to risk their health to work, especially the front-line workers. It must've been a long and exhausting year for them. For that, I salute them and thank them for all that they've done for our society and for me to do what I can do every day.

Anyways, I wish everyone a Merry Christmas and happy holidays. Let us all hope that 2021 would be a year of improvement and a steady recovery to what we call "normal".