> Nima Pourjafar

[ projects ]

[ blog ]

The past 8 months have been pretty busy and pretty awesome! I wanted to speak a bit to my experiences and what lays ahead

For context, I had spent the first four months of the year at CentML — an AI inference startup based out of Toronto, and the next four months at Figma

I'm also releasing this on my 20th birthday. I've added a reflection at the end.

CentML

I was pretty excited to work here. I had upper year friends who had done co-ops at CentML before, and they all said it was the position they learnt the most in. I can echo these thoughts now!

I had never worked with huge mono-repos, navigating hundreds of pods through k8s, and skimming through Pulumi stack diffs. I was definitely thrown into the deep-end by my manager + mentor, but it was all in the good spirit of promoting agency.

I was mainly responsible for a lot of infrastructure + backend work.

A big thing I did was spearhead a bunch of new observability work for our model serving. CentML cares a lot about meeting a certain tokens per second SLA for customers, as well as making sure we are on the top of Artificial Analysis' leaderboard. I figured out how to setup metric pipelines from our LLM serving engines on a separate port that a Prometheus ServiceMonitor could pick up, so we could have a nice TSDB to view our analytics.

On top of that, I also got to squash some memory leaks in our python code (yes, Python can leak memory) and making more of our code thread-safe.

I slowly transitioned from a very engineering/optimization heavy role, into something of a forward-deploy engineer over time. We had some significant customers who would ask for features, and my manager was nice enough to make me responsible for some of them. I would have to go from idea -> research -> demo -> production in less than a week, but it was pretty fun getting to show the end result to customers.

I would say the main things that categorized CentML's work was: agency, moving fast, customer-driven development, and knowing how to build systems that would scale.

I really enjoyed it, and it was pretty obvious CentML was going to go somewhere. Which luckily enough was a nice acquisition from Nvidia. I should've taken up my manager's offer to drop out and work full-time for them 😔

Winter Shenanigans I was living with my parents to save money and commuting to work. It's really weird driving past your old high school, sleeping in the room you grew up, and seeing faces from a distant past — all while realizing I wasn’t the same person I had been before college

I would occasionally go back to Waterloo to visit friends and work on some things. A big thing was Matia and I's severance fan project — we recreated the macro-data-refinement in the show as a collaborative web game. It offered some cool challenges to make: player state management over sockets, virtualization of the grid, and more. The project went semi-viral on Russian facebook and got some cool users. We will put it back up when season 3 airs

I really enjoyed this term. It felt like a good balance of challenging work + hanging out with friends. I was irked by a couple things though

  1. I felt like I had to get pushed outside of my comfort zone. CentML felt familiar after shipping some code, I was living in my home-town. I wanted something with a bit more entropy
  2. As much as I would go back to Waterloo, and I have great friends there — I wanted to be surrounded by smart young people every hour of the day
  3. It was cold

Lucky for me, 5 days after I had ended at CentML, I was flying straight to San Francisco to fix all these and start a new chapter at Figma

Figma

Figma’s internship program (like many at larger companies) is very structured. They make sure interns always have clarity on their progress, tasks, and overall direction. Projects are planned out in advance, check-in meetings are scheduled, and teams are thoughtfully chosen to ensure interns feel comfortable and supported.

This is all part of a bigger picture of being able to on-ramp easily to maximize performance for a full-time return offer, and climbing up the tech corporate ladder.

For what its worth, I do not care about this. I am only 19 (at least at the time of typing this up) and I would much rather be giving ambiguous, hard, and "moon-shot" tasks (obviously working at a saas company is not very moonshot, but let a man dream).

I really made it my task to get through my intern tasks project as fast and effectively as possible — so my team would trust me as if I was a normal contributor and I could be given more senior work.

The invariant of any work is that the challenge tapers are you get deeper and deeper

I think the goal ultimately is not to continue riding this tapered challenge curve, but to start new challenges that can kick start a new jump in learning

(the blue line is if you were to start anew)

My manager, mentor, and team at Figma really enabled me to do that — I got to complete four different main features in the time I was there + some misc tasks.

Some main ones include

While all of this had a significant impact and delivered real value to shareholders (all of these features were directly requested by customers and users), at a company the size of Figma, you don’t really get to influence the product direction or vision — and most of the hardest problems have already been solved. Being a regular contributor to something as big as Figma is definitely a valuable experience and a high-signal credential, but I often found myself wanting a bit more. While I was there, I learned a ton about how Evan Wallace architected the initial WebGL rendering logic and multiplayer CRDTs.

There are three types of fun:

I think you can make a bijection between these types of fun and the types of work engineers do:

In general, I think naive vibe-coding (getting Cursor to just complete the task in one-pass) covers most of type 1 engineering. Some of these coding agents that have been RL-ed on and have some planning step can complete Type 1.5 problems.

All that said, as an intern at most companies, you’re likely to be limited to Type 1 engineering. It’s the most valuable use of your time from your employer’s perspective — they won’t have you spend a week architecting a new solution for a feature that’s not even on the roadmap. With limited time, it makes sense to focus you on tasks that are well-defined and can be completed reliably.

With this, I’ve developed a new point of view that I’m applying to all the work I do moving forward.

You should always maintain an “intern” mentality — constantly curious, always learning from the work you do, and building intuition about best practices, design systems, and project lifecycles.

At the same time, you shouldn’t limit yourself to “intern” work. Don’t settle for the easy, well-trodden tasks. Instead, seek out problems that no one else has solved, take on challenges that push you, and make a deliberate effort to create impact. That’s how you grow, contribute meaningfully, and build expertise that lasts.

I take on “internships” because I believe that learning is one of the most important purposes in life. Much like Richard Sutton, I see humanity’s progress as driven by exploration, experience, and adoption — and I believe it is each person’s responsibility to carry this cycle forward, ensuring that knowledge, insight, and innovation continue to grow and benefit those who come after us (AGI, aliens, whatever it may be).

But an "internship" does not have to be something that explicitly says it is an internship. I am open to taking a gap-year to do my own independent research/learning. I'm open to dropping out to work full-time somewhere that enables this. And I'm hoping I can get to more projects I want to tackle that can push this narrative forward.

Kleiner Perkins Fellowship

I was also a Kleiner Perkins Fellow this summer. It is essentially a program that takes interns at a bunch of KP portfolio companies and organizes events for the cohort to hang out with.

This was something very fun, and I did make some very good friends out of this fellowship. Some of the operating folks at KP are some serious veterans as well, so it was super awesome getting to know them.

But I think what I became most aware of is how generic it’s become to be an intern in the Bay Area over the summer — I hadn’t realized I was part of a pipeline that reflects the growing MBA-fication of studying computer science.

MBA-ficiation of CS Degrees

I say this as a personal discovery. To me, SF is this quirky strange place where people come to go heads-down and work on things with asymmetrical upside. I've crashed at several hacker houses and friends places, to be surrounded by some of the most interesting work going around.

My experiences this summer highlighted that a lot of this is now more accessible to more undergraduates, and it does not have that same quaint charm anymore. Talking to YC founders 4-5 years ago is completely different now. Especially now that they have announced early decision — it seems like something that was uniquely intra-personally driven like YC has now become another pipeline.

In my opinion, building intellectual capital no longer is viable if you are a CS undergraduate prioritizing employment. Building social capital is now the ultimate asset — it always has been, but now it is explicit.

I think this shift is a natural consequence of how institutions evolve once they become aspirational. How what was something once for the 90s kids, is now something for everyone. The era of hackers being nerdy and weird archetypes is done. What was once countercultural becomes careerist. Hacker houses become feeder programs. “Building something cool” becomes indistinguishable from “doing something that signals you’re a builder.” It’s not necessarily bad — but it changes the texture of the ecosystem.

I started realizing that I wasn’t immune to this either. The incentives are subtle but powerful — you start optimizing for what looks impressive rather than what feels interesting. Somewhere along the line, the joy of building gets abstracted into a portfolio of career moves. Scrolling salaries, debating in threads, and gaming hypotheticals — normal behavior that creates abnormal values.

Pushing 20 (becoming unc)

I have this silly thing I say whenever someone asks who my idol is: “myself in five years.”

I remember being fifteen and having to explain that answer in an interview. I’d close my eyes and picture the version of me I thought I wanted to become — the one who would skip the ordinary steps, start working right away, drop out, and do something huge. Whatever I was doing then, I wanted ten times more of it.

But none of that really happened. I’m not working on something elite. I’m in college. I haven’t done anything monumental yet. And as I type this from the same childhood home where I once gave that answer, I can’t help but feel that my younger self might be disappointed.

Yet, somehow, I’m the happiest, most grounded, and most secure version of myself I’ve ever been.

I was raised to be agentic — to bend the world to my will, to shape outcomes through sheer force of intent. To craft a narrative and live a life worthy of being told.

While all of that still feels true, I’ve come to see agency differently. Its greatest gift isn’t control — it’s clarity. It’s the ability to rationalize what you care about, to step willingly into the fire and find a reason to stay there, even when it burns.

It’s a selfish thing, in a way. You lose people, comforts, and fragments of who you once were by standing in it — but in return, you come closer to yourself. Muslims perform wudu before prayer; Jews immerse themselves in the mikveh — both acts of cleansing, stripping yourself bare, and testing what still matters once everything else is gone.

We’ve all had to step into different fires in our own ways — moments of transition that pull us closer to who we are. They’re often sparked not by others, but by our own questions, our own choices. The solo trip you take on a whim, the book you start reading for no reason you can quite name — unprompted, unrational, but undeniably yours.

I can name all of mine — every fire I’ve stepped into, every piece of my heart that was lost in the process. But I can also feel the parts that grew stronger, the ones reborn in their place. We burn things with fire, yes — but we also kindle new ones with it.

I’ve been obsessively thinking about what the next fire I want to step into is — what role I want to take in this world. Is it better to help run it, or to simply live in it?

I feel a pull toward building, toward amassing capital and reshaping things around me to reflect my values — because so much of what governs us feels flawed, misaligned, overdue for change. But I also feel a pull toward living — to seeing the world as it is. To feel what an eclipse looks like. To talk to strangers and feel the small joy of learning something new. Sometimes that’s enough to fill the hunger for experience.

And yet, there’s an urgency that shadows all of it. The inevitability of AGI makes me feel like I should throw myself into any fire I can, before they’re all extinguished by the singularity.

Either path is selfish — and both are terrifying. But maybe that’s the nature of commitment: to walk toward the flame knowing it might burn, and to have a reason good enough to stay.

Ultimately, when we stand on the podium of judgment — whether that’s divine, societal, the quiet reckoning with ourselves, or death — what will matter isn’t which fire we chose, but whether we stepped into it willingly. Whether we lived with intention, and let it shape us into something true.

If you read all of this, it means a lot! Feel free to reach out, and I'll post some new exciting updates soon!