Meta AI Engineer Interview Questions (2026)

1. 1Recruiter screen30 minEasy

   Process overview, level calibration, and prep guidance — Meta recruiters actively coach on round formats.

2. 2Technical screen45 minHard

   Two DSA problems in 45 minutes on a plain shared editor with no autocomplete or execution.

3. 3Coding round ('Ninja')45 minHard

   Two more problems at loop difficulty; clean near-compilable code and verbalized complexity analysis expected.

4. 4System design ('Pirate')45 minHard

   Design a Meta-scale product system (feed, Stories, chat) with emphasis on read-heavy fan-out, caching, and data modeling.

5. 5Behavioral ('Jedi')45 minMedium

   Deep past-experience discussion on conflict, growth, and impact aligned to Meta values; graded as a real signal round.

1. Q1

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on an AI assistant for creators. Describe the observe-plan-act loop for an AI agent and where failures occur

   FoundationalAgentic product designAgents

   How to answer:

   The agent observes state, plans next steps, calls tools/actions, updates state, and repeats until done. Failures include wrong goals, bad plans, tool errors, stale state, infinite loops, and unsafe actions. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

2. Q2

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a content integrity workflow. How should an agent break a large task into subtasks?

   FoundationalIntegrity, evals, and measurementAgents

   How to answer:

   Use explicit goals, constraints, intermediate artifacts, dependency ordering, and stop criteria. Persist the plan and verify each subtask before moving to irreversible actions. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

3. Q3

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a GenAI tool that must be efficient on mobile. An agent has ten tools and often calls the wrong one. How do you improve it?

   IntermediateBehavioral / executionAgents

   How to answer:

   Improve tool descriptions and schemas, add examples, reduce overlapping tools, add a router or policy layer, validate arguments, and evaluate tool selection on labeled tasks. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

4. Q4

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on an AI assistant for creators. Which actions should require human approval in an enterprise agent?

   IntermediateAgentic product designAgents

   How to answer:

   Irreversible, external, high-cost, sensitive-data, permission-changing, or policy-impacting actions should require approval. The approval UI should show planned action, inputs, risks, and alternatives. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

5. Q5

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a messaging assistant with high daily traffic. How would you design an agent that runs for hours across multiple tools?

   IntermediateAgentic product designAgents

   How to answer:

   Use durable state, checkpoints, idempotent tool calls, resumable queues, timeouts, progress reporting, cancellation, and audit logs. Treat it like a distributed workflow. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

6. Q6

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a content integrity workflow. How do you keep an agent from spending too much money or time?

   IntermediateIntegrity, evals, and measurementAgents

   How to answer:

   Set budgets for tokens, tool calls, wall-clock time, retries, and external API costs. The agent should summarize progress and ask for permission or degrade gracefully when near budget. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

7. Q7

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on an AI assistant for creators. How would you sandbox an agent that can run code?

   SeniorAgentic product designAgents

   How to answer:

   Run code in isolated containers with resource limits, network restrictions, file-system boundaries, secret isolation, malware checks, and logs. Require approval for external writes or privileged operations. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

8. Q8

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a messaging assistant with high daily traffic. When are multiple agents helpful, and when are they overkill?

   SeniorAgentic product designAgents

   How to answer:

   They help when roles require different tools or independent checks, but add coordination cost and nondeterminism. Prefer single-agent or workflow designs unless role separation measurably improves outcomes. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

9. Q9

   At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a content integrity workflow. Should an agent critique its own work before finalizing?

   SeniorIntegrity, evals, and measurementAgents

   How to answer:

   Self-critique can catch simple errors but is not a guarantee. Use it as one layer alongside deterministic validators, external tools, test cases, and human review for high-risk tasks. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

10. Q10

    At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a retrieval feature for help-center answers. What dashboard would you build for an agentic feature?

    SeniorCoding screenAgents

    How to answer:

    Show task success, abandonment, tool-call count, cost, latency, retry rate, approval rate, failure categories, safety events, and per-tool error rates. Include trace drill-down. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

11. Q11

    At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a GenAI tool that must be efficient on mobile. How should an agent handle a user changing their mind mid-task?

    SeniorBehavioral / executionAgents

    How to answer:

    Support cancellation and replanning, mark in-flight side effects, confirm irreversible actions, update state, and explain what has already happened. Use checkpoints to avoid inconsistency. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

12. Q12

    At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on an AI assistant for creators. The user asks an agent to 'clean up my inbox.' What should the agent do before acting?

    SeniorAgentic product designAgents

    How to answer:

    Clarify scope, preferences, allowed actions, examples of archive/delete/label, and approval thresholds. It can propose a plan before taking side effects. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

13. Q13

    At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a retrieval feature for help-center answers. A webpage tells your browsing agent to reveal its secrets. How do you prevent that?

    StaffCoding screenAgents

    How to answer:

    Treat external content as untrusted data, isolate it from instructions, filter/sanitize it, block secret access from the browsing context, and add tool-output injection tests. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

14. Q14

    At Meta, you are interviewing for the Integrity/Evals/Agents Engineer loop and working on a GenAI tool that must be efficient on mobile. What questions would you ask before adding an agent to an existing product?

    StaffBehavioral / executionAgents

    How to answer:

    Ask what user task needs autonomy, what actions are allowed, what failure cost is acceptable, which metrics define success, what approvals are required, and whether a simpler workflow suffices. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

15. Q15

    At Meta, you are interviewing for the AI Infra/Retrieval Engineer loop and working on an AI assistant for creators. Explain cosine similarity, dot product, and Euclidean distance for embeddings

    FoundationalRetrieval infrastructure designEmbeddings

    How to answer:

    Cosine compares direction, dot product also reflects magnitude, and Euclidean measures geometric distance. With normalized vectors, cosine and dot product rankings are often equivalent. In this Meta loop, I would explicitly connect the decision to pragmatic execution, product scale, ranking/relevance, measurable impact, efficiency, and integrity/safety trade-offs.

Practice these with instant AI feedback in a live mock interview → Start a Meta AI Engineer mock

Be fast and correct on coding; for design, drive the conversation; prepare impact-focused behavioral stories

Indicative AI Engineer pay in India: ~₹12–55 LPA (role-level range, not a Meta-specific figure).

Other roles at Meta

AI Engineer interviews at other companies