Meta Software Engineer Interview Guide

Meta's loop has two shapes depending on level and track. E3-E5 candidates go through a recruiter call, sometimes a CodeSignal online assessment (90 min, proctored, single problem with 4 progressive stages), a technical phone screen, then a full onsite. The onsite is typically two coding rounds, a design round (system design or product architecture depending on role), and a behavioral round. E3 and junior candidates have no system design round.

As of August 2026, some E5/E6+ candidates report skipping the recruiter call, CodeSignal OA, and traditional technical phone screen. On this path the loop starts with an AI-Native prescreener questionnaire, then a 60-minute AI-Native Screener interview, then the onsite. Some go straight from the prescreener to the Screener with no recruiter contact in between. The four prescreener questions cover how you use AI tools day to day, how you keep up with AI tooling advances, whether you have built AI-native work and the responsibility trade-offs that came with it, and how you judge the quality and reliability of AI output. The Screener is part behavioral (how you use AI in your own projects), part LeetCode. Austen McDonald's behavioral piece on talking about AI in interviews (thebehavioral.substack.com) is a framing reference for the behavioral half.

Meta has been rolling out an AI-enabled coding round that can replace one traditional coding interview at the onsite: a 60-minute session in a CoderPad environment with an AI assistant available. The round uses a multi-file IDE-like environment (file explorer, terminal, test runner, AI model dropdown) with a single multi-part problem, not two problems in 45 minutes. Multiple candidates report the AI being less helpful during the interview than in practice, likely via system-prompt modification. Traditional coding rounds favor speed plus correctness on two problems per 45-minute slot, with code execution turned off. Dynamic programming questions are banned.

E6+ adds a Leadership Assessment before the onsite and has a different structure (1 standard coding, 1 AI-assisted coding, 1 architecture, 1 design, and behavioral). E7+/M1 have only one coding round, which is AI-assisted. As of 2026, the behavioral round includes AI-usage questions. Team matching happens before the offer (changed in 2023); offers are contingent on team matching.

The interview loop

1. Recruiter screen (E3-E5) - Background, level calibration, and logistics. The recruiter shapes the loop. E5+ candidates may skip this entirely.
2. Online assessment (E3-E5) - 90 min proctored CodeSignal coding assessment with video and mic on. Single problem with 4 progressive stages. Not reported on the E5+ AI-Native path.
3. Technical phone screen (E3-E5) - 45 min, usually two coding problems in a shared editor. Easy/medium DS&A; hard questions are reserved for the onsite. Code execution is turned off. Skipped on the E5+ AI-Native path.
4. AI-Native prescreener (E5+) - Written questionnaire sent to E5/E6+ candidates. Some candidates report no recruiter call, CodeSignal OA, or traditional phone screen on this path. Four prompts, reordered and reworded from the standard Meta AI-Native form:
   1. Which AI tools, libraries, or platforms do you use in your day-to-day engineering work, and how do you use them?
   2. How do you keep current with advances in AI tooling, and what recent trends or technologies have you started using in your own work?
   3. Have you worked on projects where AI was central to the architecture or main functionality? How did you approach designing the AI-native solution responsibly, and what challenges came up that you would not hit in a non-AI system?
   4. How do you evaluate the quality and reliability of AI technologies and their output? What tools or metrics do you rely on?
5. AI-Native Screener (E5+) - 60 minutes, after the prescreener answers are reviewed. Part behavioral: how you use AI on your projects, with probes on specific tools and decisions. Part LeetCode: a standard DS&A problem in a shared editor. The behavioral half uses the same framing Austen McDonald lays out in his thebehavioral.substack.com piece. Talk about real AI use, not abstractions.
6. Onsite loop - Varies by level:
   • Coding (E3-E5) - Two coding rounds (one may be the AI-enabled format). Traditional rounds: two easy/medium problems per 45 min, code execution off. DP questions are banned.
   • AI-enabled coding - 60 min, multi-file IDE environment, single multi-part problem (bug fix, core implementation, optimization). Available models: GPT-4o mini, GPT-5, Claude variants, Gemini, Llama. Meta sends a practice CoderPad link beforehand.
   • System design / Product architecture - Required from E4 up (E3 has no design round). System design for infra, product architecture for product roles.
   • Behavioral - Led by the 'loop lead' who aggregates all feedback. As of 2026, includes AI-usage questions. Failing the behavioral round is an automatic no-hire.
7. E6+ Leadership Assessment - For E6 and above, a Leadership Assessment interview between the technical phone screen (or AI-Native Screener on the E5+ path) and the onsite. E6 onsite = 1 standard coding, 1 AI-assisted coding, 1 architecture, 1 design, and behavioral. E7+/M1 = only 1 coding round (AI-assisted).
8. Team matching (pre-offer) - Since 2023, team matching happens before the offer. Offers are contingent on team matching completing successfully.

What they look for

• Fast, correct coding: two easy/medium problems per 45-minute round (hard reserved for onsite)
• Design depth: distributed systems for infra roles, API/client design for product roles; no design for E3
• Behavioral signal mapped to Meta's values (move fast, impact). The loop lead aggregates all feedback
• If you draw the AI-enabled round, practice directing an assistant without losing your own reasoning; the AI may be intentionally nerfed
• For E5/E6+: concrete prescreener answers. Name the tools, name the metrics, describe the responsible-design trade-offs you made

How to prepare

• Practice solving two easy/medium problems in 45 minutes, talking throughout. Code execution is off in traditional rounds.
• Know which design flavor your role uses: system design vs product architecture. E3 has no design round.
• If you draw the AI-enabled round, practice in Meta's provided CoderPad environment beforehand. Multiple candidates report the AI being nerfed during the interview.
• DP questions are banned. Don't waste prep time on dynamic programming.
• Prepare AI-usage behavioral stories. As of 2026, expect at least one question about how you work with AI.
• For the AI-Native path (E5+): the prescreener is your entry point. Write specific answers: which tools, which metrics, which design trade-offs. The Screener probes the same ground live, so have one or two real AI-native projects ready to discuss in depth. Austen McDonald's behavioral piece (thebehavioral.substack.com) is a prep reference for framing AI use in behavioral answers.