Meta's loop is built around one constraint that shapes everything else: speed. Where Google gives you 45 minutes to reason carefully through a single problem, Meta expects two problems in the same window, and the difference is not a detail. It is the single most common reason strong engineers fail a Meta coding round. The rest of the process follows the same pattern of being shorter and more compressed than its peers: fewer rounds, faster decisions, and team matching pushed after the offer rather than before it. This guide covers each stage, the signals each round scores, how E3 through E6 are actually decided, and how to prepare for a loop that rewards fluency over deliberation.
The Full Pipeline, Stage by Stage
Application or referral
A referral gets your CV in front of a recruiter faster and gives your application a human advocate. It does not change the bar in any round. Meta recruiters do read cold applications, but referral volume dominates for engineering roles.
Recruiter screen: 30 minutes
Background, target level, location, and timeline. Ask three things: which level the loop is calibrated for, whether it includes a system design round, and whether the loop is one day or split. All three change how you prepare.
Technical screen: 45 minutes
One or two coding problems in a shared editor, usually easy-to-medium. Some candidates get a single medium problem with a follow-up extension instead. Passing here means a working solution, correct complexity, and visible pace.
The onsite loop: 4 rounds
Two coding rounds, one behavioral round, and either a system design round (E5+) or a third coding round (E3/E4). Frequently run virtually and usually back to back on one day. As of 2026 one of the coding rounds may be the 60-minute AI-enabled format described below.
Debrief
Interviewers submit written feedback and a hire recommendation, then discuss as a group. Unlike Google, the people who interviewed you are in the room for the decision, which is part of why it resolves faster.
Offer at a level
You are offered a level, not a team. Compensation follows the level band. If your loop landed between two levels, this is where a down-level offer appears, and where you can ask what evidence would have supported the higher one.
Team matching
After the offer. Conversations with hiring managers who have headcount, oriented around what you want to work on. Because the offer already exists, this stage carries far less anxiety than Google's equivalent.
Why Meta finishes faster than Google
Two structural differences. First, the debrief includes your actual interviewers rather than a separate committee reading a document, which removes a queue. Second, team matching happens after the offer, so a lack of open headcount does not stall your candidacy. Three to six weeks end to end is normal, and a same-week decision after the onsite is not unusual.
The Coding Rounds: Two Problems, 45 Minutes
Meta's coding rounds, internally called Ninja rounds, are conventional data structures and algorithms. The problems are not harder than Google's; the time pressure is. Budget roughly 15 to 18 minutes per problem including clarification, coding, and complexity analysis.
Very common
- Arrays and strings, two pointers
- Hash maps and frequency counting
- Trees: traversal, path problems, BST
- Graphs: BFS, DFS, connected components
- Intervals and sorting
- Binary search
Common
- Heaps and top-k
- Sliding window
- Dynamic programming, one dimension
- Linked list manipulation
- Stacks and monotonic stacks
- Matrix traversal
Rare
- Advanced graph algorithms
- Segment trees and tries beyond basics
- Heavy mathematics
- Multi-hour design puzzles
What separates a pass from a fail in these rounds, ranked by how often it decides the outcome:
- Pace. Finishing one problem beautifully and never reaching the second is a fail in most Meta loops. Practise with a timer set to 18 minutes per problem until that pace is normal rather than stressful.
- Clean code on the first pass. There is no compiler and no autocomplete. Time spent fixing syntax is time you do not have. Write in a plain text editor during practice, not an IDE, so this discipline is trained rather than hoped for.
- Short, decisive clarification. Two or three targeted questions about input size, duplicates, and empty input. Then start. A five-minute requirements discussion is affordable at Google and is not affordable here.
- Complexity stated without prompting. Time and space, out loud, immediately after you finish. Waiting to be asked reads as uncertainty.
- One quick trace, not an exhaustive test. Walk a single small input plus one edge case. Full manual testing of both problems does not fit in the window; interviewers know this and expect judgment about what to verify.
One practical note on the shared editor: it will not run your code. Candidates used to relying on a test run to catch off-by-one errors lose disproportionately here, because the bug surfaces during the interviewer's read rather than during yours.
One of these rounds may now be AI-enabled
Since October 2025 Meta has been replacing one onsite coding round with a 60-minute AI-enabled format: a multi-file project in a CoderPad environment with a built-in AI assistant you are expected to use. It is rolling out across SWE roles through 2026, and the preparation for it is materially different from the round described above. We cover the format, the rules, and how it is scored in our guide to Meta's AI-enabled coding interview.
The Behavioral Round
Meta's behavioral round, the Jedi round, covers past projects, collaboration, conflict, and ambiguity. Meta does not publish a named rubric the way Amazon publishes its Leadership Principles, but the themes interviewers probe are consistent, and they map closely to Meta's stated engineering culture of moving fast and owning outcomes end to end.
The project you are proudest of, in technical depth
Expect drill-downs into the decisions you personally made, the alternatives you rejected, and what you would change now. Vague ownership is the fastest way to lose this round, because interviewers push until they find the boundary of what you actually did.
Shipping under time pressure
A situation where you traded scope, quality, or elegance for speed, and how you decided. Meta wants to hear a deliberate tradeoff with a named cost, not heroics. "We cut the admin dashboard to hit the launch and shipped it two weeks later" is the shape of a good answer.
Conflict with a teammate or another team
How you surfaced the disagreement, what you changed your mind about, and what happened afterwards. The follow-up is usually about the relationship after the conflict, so include it.
Working without clear direction
A goal that was underspecified. Show the narrowing: the questions you asked, the assumption you wrote down, the smallest thing you shipped to learn something real.
A failure and what changed because of it
Owned plainly, with a specific downstream change in how you work. Deflected blame is read as a maturity problem regardless of the technical content.
Feedback you received and acted on
Concrete, ideally uncomfortable, and with evidence that the behaviour actually changed. Generic self-improvement answers score nothing.
Eight to ten STAR-structured stories cover this comfortably, because one strong project can answer four of the themes above depending on which part you emphasise. Our behavioral interview guide covers the structure and the two-minute delivery target in depth.
System Design and the E4 / E5 Line
The system design round appears from E5 upward, and it is usually the round that sets your level. Two candidates with identical coding performance separate here: the one who scopes requirements, estimates scale with real numbers, justifies each component, and names their own bottleneck lands E5, while the one who draws a correct but unexamined diagram lands E4.
Meta's version of the round leans toward product-shaped systems (a news feed, a notification service, a chat backend, a metrics pipeline) rather than pure infrastructure puzzles. That means requirements clarification carries real weight, because the product surface genuinely is ambiguous until you pin it down. The five-phase structure in our system design framework applies directly, with one adjustment: spend the first eight minutes on requirements and scale rather than five, and state which features you are explicitly not designing.
Argue for your level before the loop, not after
Ask your recruiter at the screen which level the loop is calibrated to. Loops are calibrated before they run, and the questions you are asked follow from that calibration. If the target is below where you believe you sit, make the case then, because reopening it after the debrief means asking a group to overturn a decision they already documented, which almost never succeeds.
Meta vs Google vs Amazon: Where Preparation Differs
| Meta | Amazon | ||
|---|---|---|---|
| Coding load | Two problems per 45 min | One problem per 45 min | One to two, plus behavioral |
| Who decides | Debrief with your interviewers | Hiring committee, never meets you | Loop debrief, Bar Raiser veto |
| Rubric | Unpublished, culture-aligned themes | Four broad attributes | 16 named Leadership Principles |
| Behavioral weight | One dedicated round | One dedicated round | Scored in every round |
| Team matching | After the offer | Before the offer | Team-specific from the start |
| Timeline | Three to six weeks | Six to ten weeks | Two to six weeks |
| What most often fails candidates | Coding speed | Silent reasoning | Thin, undrilled stories |
If you are interviewing at more than one of these, the coding preparation transfers almost entirely and the round strategy does not. Read our Google process guide and Amazon Leadership Principles guide alongside this one so you know which mode you are in on the day.
A Realistic Six-Week Preparation Plan
Weeks 1–2: pattern fluency
Work the very common column until recognition is automatic. Review every problem you failed and be able to say why the correct approach was correct. Volume matters less than that review step.
Week 3: speed conversion
Switch to paired problems in a plain text editor, timed at 18 minutes each, no autocomplete and no test runner. This week exists solely to convert correctness into pace, and it is the week most candidates skip.
Week 4: system design, if targeting E5 or above
Ten to twelve product-shaped prompts, out loud, timed to 45 minutes. Repeat three of them a second time, since the second pass is where structure becomes reflex.
Week 5: behavioral stories
Write eight to ten STAR stories weighted toward ownership, speed-versus-quality tradeoffs, and conflict. Rehearse them spoken and timed to two minutes, then have someone drill into each one three levels deep.
Week 6: full mock loops
Two complete four-round loops back to back in a single sitting. Interview stamina is a real and separately trainable constraint, and Meta's compressed format punishes fatigue in the third and fourth rounds specifically.
Final days: taper
Light pattern review, confirm logistics and your editor setup, and sleep. Cramming degrades exactly the fluency this loop measures.
Most of that plan is spoken and timed rather than read, because both failure modes at Meta, slow coding and thin stories, only surface under a clock with someone listening. That is the practical case for unlimited AI practice: the same problem, timed, narrated, as many times as it takes, at any hour. We have compared AI and human coaching honestly, including where a human coach still wins, and our FAANG preparation roadmap covers sequencing coding, design, and behavioral work without burning out.
Train for pace, not just correctness
Amigo runs unlimited timed practice from your resume and target level, then supports you live during the real interview with structured answers streamed in real time.
Try Amigo free →Frequently Asked Questions
How long does Meta's interview process take?
Typically three to six weeks from recruiter screen to offer, which is faster than Google. The onsite loop usually runs on a single day or across two adjacent days, and the debrief decision commonly lands within one to two weeks after it. Team matching at Meta happens after the offer rather than before, which is why the timeline is shorter.
How many rounds are in a Meta onsite?
Four rounds of about 45 minutes: two coding rounds, one behavioral round Meta internally calls Jedi, and one round that depends on your level: system design for E5 and above, or a third coding round for E3 and E4.
What are Ninja and Jedi rounds at Meta?
Ninja is Meta's internal name for the coding rounds, which test data structures and algorithms with two problems in 45 minutes. Jedi is the behavioral and design-sense round covering past projects, conflict, ambiguity, and how you work with others. The names are internal shorthand; interviewers rarely use them with candidates.
How fast do you need to code in a Meta interview?
Fast. Two medium problems in 45 minutes is the standard load, which leaves roughly 15 to 18 minutes per problem including clarification and complexity analysis. Meta's coding bar weighs speed and fluency more heavily than most FAANG loops, and this is the most common reason strong candidates fail.
What are Meta's engineering levels?
E3 is entry level, E4 is mid-level at roughly two to five years, E5 is senior, E6 is staff. Level is decided by loop performance rather than years of experience, and the system design round is what usually separates E4 from E5.
Does Meta do team matching before or after the offer?
After. Meta extends a level-based offer first, then matches you to a team, and new engineers historically went through a bootcamp period to find a fit. This is the opposite order from Google and is the main reason Meta's process finishes sooner.
Can you use your own IDE in a Meta coding interview?
No. Coding rounds are run in a shared collaborative editor with no autocomplete, no compiler, and no test runner. You should practise writing syntactically correct code in a plain editor, because losing time to syntax errors in a 45-minute two-problem round is costly.
How soon can you reapply to Meta after being rejected?
The standard cooldown is six months for most rejections, sometimes twelve for a rejection at the onsite stage. Your prior feedback is retained. Reapplying and passing is common, particularly for candidates who failed on coding speed rather than on fundamentals.
Found this useful?
Share it with someone preparing for an interview.
