The Meta SWE Interview Guide

See exactly how Meta runs its software engineer loop, from the Ninja coding rounds to Pirate system design and the Jedi behavioral, and what each round is really scoring. Then drill the questions candidates report from recent loops so the two-problems-in-forty-minutes pace never catches you off guard.

• Loop length: 6 to 12 weeks
• Expected difficulty: Medium, fast-paced
• Coding: 2 problems / round
• System design: From E5

Meta levels & pay

• Software Engineer (E3, Entry)
• Software Engineer (E4, Mid)
• Senior Engineer (E5, Senior)
• Staff Engineer (E6, Staff)
• Senior Staff (E7, Sr Staff)
• Principal (E8, Principal)

The Meta SWE interview, round by round

Recruiter screen (30 min · call)

The first conversation, usually about thirty minutes with a recruiter rather than an engineer. They go through your background, ask what you have actually built and owned recently, and work out which level to put you forward for. That level decision matters more at Meta than at most companies, because it decides whether a system design round appears in your loop at all. They will also tell you the shape of your pipeline, the rough timeline, and whether you are getting an online assessment.

• Your level is set in this call. An E4 loop and an E5 loop are different loops with different rounds and a different bar. Be specific about the scope of the projects you led, who else was involved, and what you personally decided, because vague answers here tend to get you slotted lower.
• Ask what your loop actually contains. Pipelines vary by org and have changed over the last year. Ask whether there is an online assessment, whether one of your coding rounds is the AI-enabled format, and how many onsite rounds there are.
• Ask how the offer gets gated. Since 2023 Meta runs team matching after the loop and makes the offer contingent on it, so ask how many hiring manager calls to expect and roughly how long matching is running for your level right now. That number is the difference between a five week process and a twelve week one.

Online assessment (~90 min · CodeSignal)

Not every candidate gets this stage, so ask rather than assume. It shows up most often for university and new grad pipelines and for some experienced hiring waves. Recent reports describe a timed CodeSignal session of roughly seventy to ninety minutes, proctored with your camera and microphone on, built around one larger problem that unlocks in four progressive stages rather than four unrelated puzzles. Reported examples include building an in-memory key value store or a small cloud file storage service, where each stage adds requirements on top of the code you already wrote. Everything is graded automatically, so nothing you were thinking counts, only what the hidden test cases say.

• Automatic grading gives no partial credit. Nobody reads your comments or your reasoning. Get to a working solution that passes the hidden cases, and only tidy it up if there is time left over.
• Write stage one so stage four can extend it. Because the later stages build on your earlier code, a hacked first pass costs you twice. Spend two extra minutes on names and structure at the start and you will still be able to add features at minute sixty.
• Budget your minutes per stage before you start. The usual way people lose this is sinking half the clock into one stage and never opening the last one. Set a cutoff for each stage, and when you hit it, submit what works and move.

Technical phone screen (45 min · CoderPad)

One video call with a Meta engineer while you share a CoderPad window. The standard shape is two problems in forty-five minutes, which leaves you around twenty minutes each once introductions and questions are subtracted. Most problems sit at medium difficulty and lean on arrays, strings, hash maps, trees and graphs, with breadth-first and depth-first search showing up constantly and pure dynamic programming showing up rarely. The interviewer is watching how fast you settle on an approach and whether you can keep talking while you type.

• Plan for two problems. Solving one problem elegantly and then running out of time on the second usually comes back as a borderline round. Pace is part of what is being measured, so treat twenty minutes as the budget for each.
• Say the approach and the complexity before you write code. State your plan and its time and space cost out loud. If it is the wrong direction the interviewer can redirect you in thirty seconds instead of watching you build something that will not work.
• Test your own code at the end. Walk a small input through by hand, then think about the empty case, the single-element case, and duplicates. Candidates who wait for the interviewer to find those cases give away signal they could have kept for free.

Onsite coding rounds (Ninja and the AI-enabled round) (2 rounds · 45 and 60 min)

Meta nicknames its coding rounds Ninja. From October 2025 Meta began piloting an AI-enabled round that replaces one of the two onsite coding rounds, and through 2026 it has been extending it across SWE pipelines at E6 and below, so plan for one traditional round and one with an assistant unless your recruiter says otherwise. The traditional round repeats the phone screen shape: two medium problems on a shared editor, talked through as you go. The AI-enabled round runs about sixty minutes inside a CoderPad style workspace with a file tree, an editor and a chat panel where you can talk to a model, and you work through a small multi-file codebase in phases, usually fixing a bug, then building something, then improving it. You can switch between models in the panel, but the assistant only answers in chat and cannot edit your files, so you are still the person writing and checking the code.

• Ninja rewards speed and correctness together. Two problems in roughly forty minutes means you cannot spend ten minutes deliberating before the first line. Recognize the pattern, say what it is, and start writing while you keep talking.
• In the AI round you are scored on how you direct the tool. The common failure is a vague prompt followed by pasting back an answer you have not read. Ask for one specific thing, read what comes back, and say out loud why you are keeping it or throwing it away.
• Verify everything the model hands you. Interviewers are explicitly watching for candidates who accept a confident wrong answer. Run the tests, check the edge cases yourself, and be ready to explain the logic line by line as if you had written every character of it.

System design and behavioral (Pirate and Jedi) (2 rounds · 45 min each)

The design round, nicknamed Pirate, appears from E5 upward, and E3 and E4 loops usually skip it entirely. You are often offered a choice between a system design problem, which leans toward back end and infrastructure, and a product architecture problem, which leans toward a user-facing feature with its API, data model and client behaviour. Either way you are expected to set the scope yourself, ask about scale, and commit to decisions rather than waiting to be told what to build. The behavioral round is called Jedi and is usually run by the person leading your loop, who also gathers everyone else's written feedback. Meta looks for a fairly specific set of things in it: resolving conflict, driving results, working through ambiguity, growing over time, and communicating clearly.

• Drive the design round from the first minute. State the requirements you are assuming, pick a direction, and defend it. Interviewers push repeatedly on scale and will ask you for actual throughput and storage arithmetic, so do the numbers on the board.
• Choose the design flavour that matches your real work. If your last few years have been product surfaces, a product architecture problem will show you at your best. Ask before the round which options exist for your loop, because the choice is not always offered.
• The behavioral round can move your level. Jedi carries real weight in the level decision, and a thin round can turn an E5 loop into an E4 offer even when the coding went well. Bring stories where you personally made the call, with the outcome and a number attached.

Debrief and hiring review (1 to 2 weeks after the loop)

Every interviewer writes up their round and gives a rating on a fixed scale before seeing anyone else's notes. The loop lead then assembles the packet and it goes to a review that decides two separate things: whether to hire you, and at what level. Consistency across the loop is what carries this stage. One very strong round rarely rescues a loop that also had a weak or contradictory one, and mixed signal frequently turns into an offer one rung down rather than a rejection.

• Ratings are written independently, before anyone talks. There is no chance to recover a bad round by charming a later interviewer, because their notes are already filed. Treat each round as if it were the whole decision, because for the weakest one it more or less is.
• Level is decided here, not in the recruiter call. The review can approve you at a rung below the one you interviewed for. If a downlevel comes back, ask your recruiter which round drove it before you accept or push back, because that answer tells you whether it is negotiable.
• One to two weeks of silence is normal. Assembling the packet and finding a debrief slot takes real time, and a quiet week usually means calendars rather than doubt. Ask for a target date instead of trying to read the delay.

Team matching and offer (2 to 6 weeks · variable)

Meta decides to hire you before it decides where you will sit. Once your packet clears, your recruiter circulates it to teams with open headcount and you take calls with the hiring managers who are interested, commonly three to five of them at forty-five to sixty minutes each, though some candidates report more than ten. Both sides have to opt in, and since 2023 the offer is contingent on a match, so the formal letter usually only gets generated once one of those conversations lands. That makes an approved loop something short of an offer in hand. This is the step that stretches Meta timelines the most, and it runs anywhere from a couple of weeks to well over a month depending on headcount, your level, and how narrow your interests are.

• Approved is not the same as offered. You can clear the bar and still wait weeks for a team, and in tight headcount periods some candidates sit in matching longer than the loop itself took. Keep other processes warm until you have signed something.
• Treat the manager calls as a real evaluation of them. Ask about on-call load, what the team is measured on, what shipped in the last six months, and what the roadmap looks like. The team you land on shapes your first two years far more than the company name does.
• Being very specific costs you time. A narrow team or product preference lengthens matching, sometimes by weeks. Work out in advance which of your constraints are genuine and which are preferences you would trade for a faster close.

How to prepare

Build a daily coding rhythm (Weeks 1 to 4)

Code every day for the month. Rotate through the patterns Meta leans on, arrays and hashing, trees and graphs, BFS and DFS, intervals, and heaps, starting at medium and always timed on a shared editor so speed becomes automatic.

Drill two problems per round (Weeks 2 to 4)

Meta's Ninja rounds expect two full solutions in roughly forty minutes, so practice that pace directly. Sit paired mediums back to back on a twenty-minute clock each, clarifying, coding, and testing both, until finishing two clean problems in a single round feels normal rather than rushed.

Layer in Pirate system design (Weeks 2 to 3)

For E5 and up, add system design once your coding is warm. Work through three or four full product-system designs end to end, scoping, API, data model, storage, and scaling, saying each decision aloud and rehearsing how you would drive the round and defend your trade-offs.

Prepare Jedi behavioral stories (Weeks 1 to 2)

Draft these early so you have weeks to rehearse. Write specific stories around conflict, ownership, and ambiguity, each with your exact role and a measurable outcome, and practice delivering each in about two minutes without drifting into team-level generalities.

Recently asked

Lowest common ancestor of a binary tree

About Meta

Meta Platforms builds the products a large share of the world uses to connect: Facebook, Instagram, WhatsApp, Messenger, and Threads together reach billions of people every day. Engineering hires span the core social apps, ads and monetization infrastructure, and the systems that move messages and feeds at planetary scale.

Beyond the social apps, Meta invests heavily in Reality Labs, its augmented and virtual reality division behind Quest and its work toward the metaverse, and in large-scale AI research and infrastructure. That range means engineering work at Meta stretches from consumer product velocity to some of the largest distributed systems in operation.

Frequently asked

How many rounds is the Meta interview?

Usually a recruiter screen, a technical phone screen on CoderPad, and a virtual onsite of four to five rounds: two coding, one system design for E5 and up, and one to two behavioral. A team-matching step and a committee debrief follow a passing loop.

What are the Ninja, Pirate, and Jedi rounds at Meta?

They are Meta's internal nicknames for the onsite rounds. Ninja is the coding round, Pirate is the system-design round, and Jedi is the behavioral round. Recruiters and engineers use the terms casually, so knowing them helps you map the loop.

Why does Meta ask two problems per coding round?

Meta's Ninja rounds run two problems in roughly thirty-five to forty minutes to test speed alongside correctness. It is a deliberate signal: they want to see you clarify, solve, and verify quickly, so a single slow solve, even a correct one, reads as a weaker round.

Does Meta ask system design, and from which level?

Yes, the Pirate round appears from E5 (Senior) upward. Entry and mid-level loops (E3 and E4) are coding and behavioral only. At E5 and above you are expected to drive the design and defend your trade-offs, and this round decides most senior outcomes.

What is the difference between Meta E4 and E5?

E4 is the mid-level Software Engineer, expected to own well-scoped work end to end. E5 is Senior, expected to lead ambiguous projects and influence beyond their immediate team. The clearest loop difference is that E5 adds the Pirate system-design round and raises the bar on driving decisions independently.

How does Meta team matching work?

Meta often decides to hire before placing you on a team. After a passing loop you enter team matching, where you talk with hiring managers to find a fit. Historically some hires also joined through a bootcamp and chose a team after onboarding. Your final team, and sometimes level, can settle at this stage.

What is the Meta bootcamp?

Bootcamp is an onboarding path where some new engineers ramp up before committing to a specific team, sampling projects and teams first. It has been more common for earlier-career hires and is one of the ways Meta handles team matching after an offer.

What is the Meta E5 (Senior Engineer) salary?

Total compensation for E5 averages roughly the high-300s in thousands per year, combining base near $230k with RSUs vesting evenly over four years and a one-time year-one sign-on. Use the estimator above to model your own location and level.

What behavioral signals does Meta look for?

The Jedi round focuses on conflict, ownership, and navigating ambiguity. Interviewers want concrete stories where you explain your specific role and a measurable outcome, not team-level summaries. Prepare a few sharp examples rather than one generic narrative.

What coding topics does Meta focus on?

Meta leans on core data structures and algorithms: arrays and hashing, strings, trees and graphs, BFS and DFS, intervals, heaps, and binary search. Most questions sit at LeetCode medium, with the difficulty coming from pace rather than exotic topics.

How hard is the Meta SWE interview?

Medium and fast-paced. Most questions rate medium rather than hard, but the two-problems-per-round format makes speed the real challenge. The system-design round is where senior loops are typically won or lost.

How long does the Meta interview process take?

Usually six to twelve weeks in total. The interview stages themselves are quick, often three to four weeks, but the debrief takes one to two weeks and then team matching begins. Meta matches you to a team after you pass, not before, and that step depends on which teams have headcount when you clear. It is the most common reason an offer that felt close takes another month to arrive.

How are Meta interviews scored?

Each interviewer rates you on a fixed signal scale, from Strong Hire down through Hire and below. The hiring committee reads the ratings together, so consistency across coding, design, and behavioral matters more than one standout round rescuing a weak one.

Can I reapply to Meta after a rejection?

Yes. Meta typically asks candidates to wait before reapplying, commonly around six months to a year depending on the outcome. Use the gap to close the specific gaps your loop revealed, most often coding pace or system-design depth.

What platform does Meta use for coding interviews?

Meta runs its coding rounds on CoderPad, a shared screen-sharing editor. There is no separate auto-graded online assessment for most SWE loops; the phone screen with an engineer is the first coding filter.

How should I prepare for the Meta interview?

Drill mediums daily and practice two problems per round on a clock, prepare Jedi stories around conflict and ownership, and for E5 and up rehearse driving a full Pirate design out loud. Karavine's Meta pack is exactly this plan with worked solutions.

Meta interview questions on Karavine