软件工程面试迎新变化：大模型重塑招聘标准

Our software engineering interviews are changing (a little)

3 min read

Over the last few years, LLMs have changed how software gets built and that’s no different for us at Jane Street. Most of our engineers now use them every day not only for writing code, but also for reading, debugging, and reviewing code, and plenty of things that aren’t code at all.

We’ve talked about this recently in a podcast on how we use AI in our development, and in a post on formal methods that touches on how agentic coding is reshaping the way we work.

Our coding interviews haven’t changed, though. The questions we ask software engineers today are the same style of question we’ve been asking for years: sit down with an interviewer or two, program through a problem together, and talk as you go.

That style has served us well and the key goals remain the same: we want to see how you solve problems, structure your work, communicate, and collaborate in the pursuit of excellent software development. But the shape of the job has evolved, and we think our interviews should evolve, too. As generating code has become arguably cheaper than ever, other parts of the job have come to matter even more: taste about what to build, choosing abstractions that last, and making sure what got built is actually correct, among other things.

So this summer we started trialing some new kinds of interviews in our onsites for campus candidates. Broadly, they involve less code-writing, and lean more higher-level, focusing on software design, on verifying the quality, structure, and intention of code, and on using LLMs to produce code. Most rounds will continue to be traditional coding interviews, conducted without LLM assistance.

We know this sounds vague, but that’s intentional: the formats are still evolving, and we’d rather not describe something in detail today that looks different in six months.

A few things worth knowing:

Not every onsite includes these new rounds yet, and when you come on-site we’ll give you a heads-up about what to expect.

· The rest of our process is unchanged. If you’ve read our earlier posts about

interviewing here, they’re still accurate.

· While we find LLMs incredibly valuable as part of our daily work, we also recognize that

not everyone has deep experience working with them. You shouldn’t feel like you need to have a highly worked out methodology for working with agents in order to do well in our process. In the end, we’re looking for people with great instincts for building software, and we don’t feel like we need an LLM to evaluate that.

· Different folks have integrated LLMs into their workflows in different ways and to

different degrees, and we don’t expect you to have a particular setup or to use them for everything. What we care about is that you think clearly about the tools you use: what they’re good at, where they fall short, and when to reach for something else.

· As has been the case for some time, our more experienced candidates’ interviews will

continue to be more varied, featuring a range of interview styles, from producing software to designing systems and discussing prior work.

As always, we’ll keep iterating, and we’ll write more when there’s more to say.

If you’re interested in working with us, apply on our website!

Matthew joined Jane Street in 2013 as a software engineer, and works on distributed systems infrastructure and productivity tooling. In his free time he noodles with making music.