Meta interview prep: the real loop

Band mapping: L5 here corresponds to Meta E5 (Senior Software Engineer), a direct ladder match. Recruiter explains the signal-based loop: each round captures specific 'signal' for the debrief. Coding rounds also occur but sit outside this round taxonomy. Team matching with prospective managers now typically happens before the offer is issued; bootcamp (now a few weeks) is onboarding, not team selection.

Product-flavored system design at E5: a user-facing system designed end to end. Interviewers grade on signal: problem navigation, quantitative reasoning, and pragmatic tradeoffs under time pressure.

Behavioral round graded against Meta's published competency set: resolving conflict, driving results, embracing ambiguity, growing continuously, and communicating effectively — plus impact orientation and growth from feedback. Meta's debrief vocabulary is signal-based: interviewers report evidence per axis rather than gut calls, and a failed behavioral round is decisive. Interviewers interrupt freely and jump between questions mid-story — candidates are expected to front-load actions over context. After the loop, a debrief committee hunts for holes in the combined feedback and occasionally sends the candidate back for one more interview instead of deciding.

Frequently a team-match conversation rather than evaluative; candidates may meet several managers and both sides opt in before an offer is issued. The hiring decision itself is made in a debrief/committee from the loop's combined signal.

Band mapping: L6 corresponds to Meta E6 (Staff Software Engineer). Recruiter screens for staff-shaped evidence: impact across several teams, direction-setting, and scope that wasn't handed to them. Down-level offers at E5 are common when staff signal is thin.

E6 design bar: same prompts as E5 are fair game, but graded for architecture-level judgment — cross-system effects, capacity and cost at scale, and where the org's engineers will get this wrong without guardrails. Staff loops commonly include two design rounds (system plus product/architecture); a project-retrospective round is sometimes reported in place of a coding round.

Staff behavioral signal: influence across teams, creating scope rather than receiving it, driving alignment in fast-moving and sometimes chaotic conditions. Calibration mirrors Meta's internal staff bar, where the committee — not the manager — owns the decision ('at the Staff level, your manager is just a messenger for the promotion committee'); candidates should expect written-evidence-quality stories, since documentation and writing are repeatedly cited as the staff-level visibility currency.

Leadership assessment reported for E6+ loops: role scope, cross-functional partnership, and conflict resolution — how the candidate sets direction for a problem area, handles org change and re-prioritization, and multiplies other engineers' impact.

Band mapping: L7 corresponds to Meta E7 (Senior Staff Software Engineer). Screens look for org-level evidence: directions spanning an org or pillar, executive interaction, and impact at the hundreds-of-engineers scale.

E7 design: architecture strategy for a product pillar — families of systems, capacity and cost at company scale, multi-year evolution, and the org design implied by the architecture.

Org-level leadership: setting direction for hundreds of engineers, partnering with directors on strategy, making big bets quickly with incomplete information, owning visible failures.

Senior-staff behavioral signal: influence at org boundaries, handling public failure, sustaining pace and culture through hypergrowth or contraction.

Charter conversation: the pillar's hardest problems and the candidate's thesis. Heavily mutual at this band.

Band mapping: L5 corresponds to Meta E5 ML engineer. Recruiter confirms MLE versus research-scientist track. ML coding rounds also occur outside this taxonomy; the ML system design round is the distinctive element. Official: the ML full loop is 'up to six 45-minute conversations' with engineers; practitioner accounts of senior MLE loops show two coding rounds, one behavioral, and two design rounds (one generic system design, one ML system design) after a two-question coding phone screen.

ML system design centered on large-scale personalization: ranking, recommendations, integrity. Signal axes include problem navigation, modeling pragmatism, and feedback-loop awareness at billion-user scale.

Impact orientation applied to ML: experiment-driven stories, metric honesty, iteration speed. Same signal-based debrief vocabulary as SWE. Practitioner hiring-manager guides add data-literacy and evaluation probes here: a misleading summary statistic the candidate caught, unexpected or biased model outputs and the guardrails around them, and what they do when performance breaches a threshold.

Typically post-loop, pre-offer matching: ranking versus integrity versus content understanding teams differ greatly in data and pace; conversation centers on fit, with both sides opting in before the offer.

Band mapping: L6 corresponds to Meta E6 ML engineer. Recruiter looks for staff ML signal: owning a model area across teams, objective and value-model design authority, experiment governance.

Staff ML design: multi-model ecosystems — ranking plus integrity plus value modeling — including the human and organizational dynamics of many teams shipping into one surface. Public Meta-specific guidance contrasts levels inside the same ML system design round: junior/mid answers center on ML algorithms; senior/staff answers are expected to drive operational coordination — data freshness, training-serving skew, rollout safety via experimentation frameworks, graceful degradation, observability.

Staff behavioral signal in ML: cross-team objective disputes, experiment quality governance, creating model-area scope rather than inheriting it.

ML technical leadership: setting modeling direction for a surface or domain, balancing engagement with long-term user value, growing senior MLEs.

Band mapping: L7 corresponds to Meta E7 ML engineer. Screens for org-level ML evidence: modeling strategy for a pillar, ranking/value frameworks adopted across surfaces, executive-level AI partnership.

E7 ML design: pillar-scale ML architecture — shared foundation-model strategy across surfaces, compute economics, unified value modeling, governance of dozens of model teams.

Org-level ML leadership: multi-year modeling bets, paradigm-shift navigation, balancing engagement economics with societal and regulatory pressure.

Senior-staff behavioral: high-visibility failure ownership, candor with executives about ML limits, sustaining research-to-production pipelines through org change.

Charter and thesis conversation for a pillar's ML direction; heavily mutual at this band.