The Value of Niches

I’ve hesitated to write this post. Not because it’d be controversial or because it’d get me in trouble. But because I’m concerned that people might misconstrue or exaggerate the message.

Let me be unambiguous here:

You should keep learning all aspects of computer science

You shouldn’t feel obligated to stick to a particular area

There is no one magical area that will get you hired. No, not even ML.

With that out of the way, let’s get on with the post.

I’m interested in programming languages and compilers. I find them really fascinating for a variety of reasons. They’re a foundational aspect of programming; they influence how people think about their problems; and they have a host of cool engineering challenges.

I became interested in programming languages in my freshman year of college. I stumbled upon Crafting Interpreters, started coding along and got hooked. Since then I’ve attended workshops on programming languages, tried (and kinda failed) to work on the Ruby interpreter, (slowly) built my own compiler and finally got a job working on a compiler. In short, I found a niche.

My niche has helped me a lot. I’ve gotten good at an area, moreso than your average undergrad. I’ve managed to connect with people in my field and learn from them. And the benefit you’ve all been waiting for, I’ve managed to get jobs out of it.

It becomes a lot easier to sell yourself when you can confidently say that you are in the few hundred or so who know this area. I can debug WebAssembly by reading the hex. I’ve read the WebAssembly spec and understand how it does first class functions. I understand the nitty gritty bits of compilers like closure conversion.

Of course, gaining this knowledge takes work. I had to learn from different sources, work on side projects, bang my head against the wall (proverbally), etc. But it’s not an impossible amount of work. I’ve met other college students who have done the same or more. It’s the focus and specificity of the work that makes it unusual.

I’ll give an example. I worked on a compiler for my language, Saber, for a while. It’s still not fully working but I managed to get a few core concepts done like strings, structs, closures, etc. I sent an email to the Cloudflare CEO explaining that I had written this compiler and that I was interested in working at Cloudflare due to their work on WebAssembly and Rust. To my surprise, he responded! And I got the job!

This may seem extraordinary, but it’s actually not. For one, CEOs respond to their email a lot more than you’d expect. I’m 2/3 for emailing billion dollar company CEOs (1/1 for trillion :D). And remember, this is a niche. There’s only a few hundred¹ people who have experience with WebAssembly and writing compilers and writing Rust. The majority of them are already in these positions. You could train a regular intern for this, but that’d likely take a few months between learning Rust, WebAssembly and compiler techniques.

Hiring someone who knows this exact niche and can start immediately is on par with saving a few thousand dollars in training, recruiting, etc.

Now niches do also limit where your experience is strictly relevant. But that’s not a huge problem. Most companies don’t hire based on what they exactly need this second. If you had a position open for a Java engineer using Spring and Android, you wouldn’t wait to hire the one person who is an open source contributor to Spring, Android and wrote a Java compiler. But, if said person did apply, you’d probably snap them up pretty quickly, no?

Think of this as a multiplier. If you happen to hit the right buttons, you’ll get a 10x boost on hireability. If you don’t, you’ll just be a normal candidate.

The other great part about having a niche is that you can demonstrate your competency by just talking. The moment I start talking about WebAssembly internals, or Hindley Milner typechecking or parser error recovery, the interviewer understands that I know my stuff. After that, they’re much more likely to give me pass on my leetcode performance.

How do you get a niche? The most natural way is to stumble upon something that seems interesting. Maybe this happens in class. Maybe it happens on reddit or Hacker News. It could be a blog post on embedded operating systems or a lecture on zero-knowledge proofs. One you have that interest, you can go off and do some research. Ask professors, read papers, skim some textbooks.

The next most organic way is to look at your existing interests and see if you can parlay them into niches. Did you like tinkering with electronics? Maybe embedded systems is your thing. Do you like dance? Maybe wearable electronics and human computer interaction is your area. Are you into a sport? Maybe you could use computer vision to analyze it.

If you can’t find a niche organically, then you can still explore different areas. First, ignore all the web/iOS/systems, whatever. That’s still too broad. Get very very very specific. Learn how one particular operating system/language/library works. Read the source code. Don’t be afraid to go low level too. If you have to trawl through hex or read assembly, so be it. Contact people in that field and learn from them.

It’s okay if you make slow progress. I’ve inched along in PL for years. I’ve gone weeks, months without touching my compiler. What’s mattered is that I came back to it and kept working. Even if it took me 3 years when someone else took 3 months, I still made it.

What a niche demonstrates is an ability to choose a topic and persist. It demonstrates that you can teach yourself, as your knowledge should go deeper than any course. It demonstrates that you find something genuinely interesting within CS and took the time to explore it. All of these are really great signs to a potential employer.

Oh also the Niche Zero is a pretty awesome espresso grinder. Not related to anything CS, just saying.