Say you are interviewing at Perplexity for an engineering role. Depending on the role, one round may forbid you from using AI to write your answer, and another may require you to use a frontier coding agent. Same company, same interview loop.
That is not a contradiction. It is a pretty clear statement of what they want to know about you, and Perplexity writes all of it down on its own interview portal (interviewresources.perplexity.ai).
One title for everyone technical
First, the part you notice on the job board: "We have one title for technical team members at Perplexity: Member of Technical Staff."
So there is no "Senior Backend Engineer" or "Staff ML Engineer" on the badge. The portal explains the spirit of it: "Engineers do data science. Designers implement features end-to-end. Even our lawyers write and analyze code."
That shows up in the postings. Of the 129 roles live on Perplexity's job board on October 6, 2026, 43 carry "Technical Staff" in the title, with the team in brackets after it (Search Core, Security, Infrastructure and so on).
Two rounds, two opposite rules about AI
Perplexity lists six interview formats: Practical Assessment, Hands-on Coding, Spike Round, Agentic Coding, System Design and Core Expertise. Two of them set opposite rules.
The Practical Assessment is the no-AI one:
"We do not allow the use of AI or other outside assistance."
You may read your language's documentation, and you may ask a search engine or AI tool about a library, but "you should not ask the tool to produce any significant part of your answer."
It is self-scheduled. "Most practical assessments are designed to be completed in 45 to 90 minutes", inside a 2 or 4 hour window.
The Agentic Coding round is the opposite:
It "require[s] candidates to use frontier agents, paired with their own judgment and expertise, to solve a technical problem."
The take-home version is "typically fixed in length at 3 hours". The live version runs "between 60 and 120 minutes".
The live one is judged like a normal coding round, "but with appropriately higher expectations for scope, depth, and quality of work."
Read together, the message is fairly plain. They want to see what you can do on your own, and separately what you can do with an agent, and they expect more from you in the second case, not less.
There is also a Spike Round, where candidates "define and then deliver a project from idea to implementation, using the tools and tech stack of their choice."
(The portal also spends real space on cheating, down to "Cheating tools running on second, third, and even fourth monitors" and a mention of "Cluely et al." by name. If you were wondering whether interviewers notice, they say they do.)
What is open, and where
Perplexity's own board had 129 live postings on October 6, 2026. By location:
| Location | Postings |
|---|---|
| San Francisco | 101 |
| Belgrade | 11 |
| New York | 8 |
| London | 6 |
| Berlin | 2 |
| Washington, DC | 1 |
So it is overwhelmingly a San Francisco company, with Belgrade a clear second, which is worth knowing if you are in the Balkans or nearby. Several postings state in-office expectations directly: one sales role asks for "the San Francisco or New York office four days per week", and a search quality role "is hybrid in Belgrade, London or Berlin."
On our board, 60 of those roles are tech roles we have on file.
Pay
109 of the 129 postings publish pay. The engineering band is notably uniform: 23 of the 43 Member of Technical Staff postings carry the identical "$220K - $405K - Offers Equity" range. One backend platform posting puts "typically 4+ years for mid-level, more for senior and staff" under that single band, which is how one title and one range can cover several levels at once.
So the top of that band is for someone far more senior than the bottom. If you get an offer, ask where in the range you are and why.
Who this is for
If you are an engineer who is comfortable both with and without an AI agent, likes the idea of one title and wide responsibilities, and can be in San Francisco (or Belgrade, London or Berlin for some teams), Perplexity has a lot open: 129 postings on its board. The interview portal is worth reading in full before you apply. Most companies make you guess at their process. This one tells you.
Its open tech roles are on Perplexity's page on Remoet. Star it and its next ones come to you.
Sources and gaps
Everything about the interview comes from Perplexity's own portal, read on October 6, 2026. Posting counts, locations and pay come from Perplexity's public job board the same day. Its careers hub would not load for us, so we have not described benefits or a pay philosophy. We have not interviewed there. If you have, especially the agentic round, tell us how it went in Discord.