Q2 2025 Update
update on what I've been working on in Q2 of 2025!
The past few months have felt quieter on the surface, but beneath it, a lot has been shifting.
I started my internship at Shopify, where I’m working on the foundational models team, exploring the space of generative recommenders and learning what it means to build at scale. I’ve been diving deeper into research under the guidance of my mentor, trading breadth for depth, iteration for rigour. I’ve also continued my work at Fallyx, developing ML models for fall detection and activity recognition, work that’s directly shaping safety outcomes for senior residents. UofT AI continues to be a meaningful through line; now from the vantage point of co-president, where I’m learning just how much leadership is about listening. I squeezed in a few trips to Alberta and Vancouver. Wrote a blog post about balance (or, more accurately, why we don’t actually want it). And have been thinking a lot about what it means to commit to fewer things, more fully.
If Q1 was about momentum, Q2 has been about calibration. Less sprint, more stride. Less noise, more attention. I’m starting to feel the difference between doing a lot and doing what matters and I’m learning that the latter usually takes longer, feels quieter, and asks more of you than you expect.
Here’s what I’ve been building, reflecting on, and beginning to understand, one quarter at a time.
research takeaways.
Over the past couple months, I’ve been working on a couple research projects at the Vector Institute, and they’ve really opened my eyes in terms of how good research should be done. I’ve been learning a lot from my research mentor, not just about methodology or models, but about mindset. About what it means to approach a problem with care, depth, and intellectual honesty.
In some ways, it was jarring. I came in with a naive mindset → quick iterations, constant updates, momentum over polish. But research runs on a different clock. There are no weekly shipping deadlines or dashboards of visible progress. Just you, the problem, and the uncomfortable ambiguity of not knowing how long it’ll take or whether you’re even heading in the right direction.
At first, I struggled with that slowness. I felt like I wasn’t “moving fast” enough. But eventually, something shifted. I started to appreciate the stillness. I started to see how many shortcuts I was unconsciously taking in other parts of my work and how research leaves no room for that. Every assumption has to be interrogated. Every result, verified. Every idea, made watertight.
You can’t fake your way through good research. You have to earn every inch of progress. And oddly, that rigour felt…liberating. Because once you give up the need to constantly appear productive, you can actually become productive in a much deeper way. You stop optimizing for output and start optimizing for understanding. You stop chasing newness and start chasing clarity.
Somewhere along the way, I realized the only metric that really matters—at least to me—is waking up excited to work on a problem. That’s it. Not the prestige. Not the publication. Just that spark of curiosity that keeps you coming back, even when the work gets hard or boring or slow. That spark is what makes depth possible.
We talk a lot about discipline and drive, but curiosity is the real engine of deep work. You can’t brute-force your way through a hard problem unless something inside you wants to stay with it. And I think that’s what research revealed to me. How much of good work is not about talent or speed, but about sustained attention. About showing up again and again with humility and care.
But maybe the more unexpected lesson I’ve taken from research is this: learning how to stay with uncertainty. Most of the time, you don’t know if you’re on the right track. Your results don’t match your expectations. Your baselines don’t improve. Your hypothesis breaks. And yet, you keep going. Not out of blind optimism, but because uncertainty isn’t a sign you’re lost. It’s a sign you’re actually exploring.
Research taught me that uncertainty is not a threat to progress, it’s the precondition for discovery. The deeper you go, the less certain things become. But that’s not a failure of understanding. That is understanding. And the more comfortable you become with not knowing, the more likely you are to find something that actually matters.
So if there’s one thing I’ve taken away from this season of research, it’s this: fast is useful, but depth is irreplaceable. Speed might get you somewhere quickly, but only depth gets you somewhere meaningful.
on working at scale @ Shopify
This summer, I joined the foundational models team at Shopify, where I’ve been working on large-scale ML infrastructure and generative recommender systems. On paper, it was a dream: the intersection of cutting-edge research and real-world impact, at a company operating at one of the largest commercial scales on the internet.
But what surprised me most wasn’t the technical complexity. It was the mental shift that working at scale requires.
At smaller orgs or in research, there’s this satisfying tightness between idea and implementation. You build, test, iterate. You can see the full system in your head. There's room for intuition, tinkering, personal ownership. But at scale, the systems are too big for any one person to hold. You can’t brute-force your way through with cleverness or sheer velocity. You have to slow down, understand the constraints, respect the architecture, and sometimes step back.
What I realized is that scale doesn’t reward ego. It rewards rigour, coordination, and a deep respect for context.
At first, this felt frustrating. I was used to shipping quickly, proving ideas through scrappy prototypes, moving fast and breaking things. But here, moving fast without understanding just meant breaking things you couldn’t always fix. The cost of a misstep was higher. The time to alignment was longer. And the bar for quality was so much higher, not because people were slower or more cautious, but because at scale, every decision ripples.
And so I started to reframe the way I approached my work. Instead of asking “how can I build this quickly?”, I began asking “how can I build this responsibly?” Instead of optimizing for cleverness, I optimized for clarity. Instead of trying to control every piece of the stack, I learned to collaborate more intentionally, to document better, to communicate assumptions, to ask more questions up front.
Working at scale made me realize that true technical maturity isn’t about how many things you can build. It’s about how many things you can integrate, sustain, and scale without chaos.
It also made me more comfortable with being small inside a large system. And that’s not a bad thing. There’s something deeply humbling about contributing one piece to something that serves millions of people. You stop needing to be the hero. You start caring more about impact than credit. You start seeing your work not as an isolated output, but as part of a larger machine that requires trust, alignment, and infrastructure to function.
There’s a different kind of creativity that emerges in that space, not the creativity of freewheeling ideas, but the creativity of working within constraints, of solving real problems with real limitations, of making things better without needing to reinvent everything from scratch.
And that, to me, is one of the most valuable lessons I’ll take with me from this summer: that scale changes not just how you build, but who you become while building. It asks for less ego, more thoughtfulness. Less sprinting, more systems thinking. And most importantly, it reminds you that good ideas aren’t enough; they have to survive in the real world, where reliability, collaboration, and context shape everything.
what you let in & what you let go.
If there’s one thing this quarter has taught me, it’s that your reality is shaped by what you let in and what you choose not to.
I used to think progress came from doing more. More projects, more reading, more people, more ideas. But I’ve started to see how much of my mental clarity and creative energy depends on being selective. The quality of my thinking is directly correlated to the quality of what I’m consuming i.e who I’m around, what I’m reading, what I’m solving for. And equally, what I’m tuning out.
This quarter, I started saying no more often. Not because I wasn’t interested, but because I wanted to be more invested in the few things that mattered. The more I narrowed my inputs, the sharper my outputs became. Fewer distractions meant deeper work. Fewer scattered efforts meant more meaningful progress. Less noise meant more signal.
I’m learning that commitment isn’t just about doing something for a long time, it’s about protecting the conditions that allow you to go deep. And often, that means letting go of things that are good, but not essential. Because even the best inputs can dilute your focus if there are too many of them.
The internet rewards breadth. But depth is what changes you.
And the more I build, the more I’m realizing: I don’t just want to collect experiences or ideas. I want to absorb them. To let them shape me. That only happens when you give them your full attention.
So going into the next quarter, I’m asking myself less often, “What else can I add?” and more often, “What’s worth fully showing up for?”
Because the inputs I choose today become the life I live tomorrow.
closing thoughts.
Looking back on this quarter, I don’t feel like I moved faster. I feel like I moved deeper. There was less noise, but more clarity. Fewer tasks, but more care. Less scrambling, more shaping.
I’m learning that real progress doesn’t always announce itself with fanfare. Sometimes, it shows up quietly, in how you think, how you focus, how you choose to spend an ordinary Tuesday afternoon. It’s not just about what you’re building out in the world, but what you’re building within yourself: taste, judgment, discipline, trust.
This season has reminded me that momentum isn’t just about doing more. It’s about doing the right things long enough for them to matter. And the more I return to that, the more aligned things start to feel.
Thanks for reading. Onward to Q3.
Until next time,
Dev
Great read Dev! I loved the part about uncertainty in research 👏