D8LooP
Start building

Interviewing

Staff Data Engineer Interview Prep for 8+ Years of Experience

Seven weeks for a loop that levels you on scope: stories, platform, governance, migrations, cost.

7 weeks94 steps~24h total
Start this plan

It opens on your dashboard and ticks itself off as you go. Every step below is a chapter or an exercise that already ships.

01

When you finish, you can

You show org-scope impact: designs other teams depend on, and stories with numbers.

Loop day Ready to be levelled Staff

02

The plan, 7 weeks

Ordered by what each one buys you, not by topic. Every step says why it is here.

  1. Week 01 · ~2h

    Scope first: the story bank that sets your level

    You can tell six org-scope stories, each with a number.

    Skip this if your stories already span teams and quantify business outcomes.

    1. ChapterBehavioral Storytelling20 minBuild six to eight stories, not twenty shallow ones.
    2. ChapterAmazon20 minAt L7, Think Big and Have Backbone decide the loop.
    3. ChapterGoogle20 minGoogleyness and Leadership often decides senior-plus hires.
    4. SectionRunning the incident and writing it up7 minStaff stories fix the process, not only the bug.
    5. InterviewTwo P1s at the same time6 minPrioritising across teams, told as a story you led.
    6. InterviewWrong numbers, one hour to the meeting6 minExecutive pressure: what you said, and what you changed after.
    7. InterviewCentral team or domain ownership6 minAn org-design question disguised as a modeling one.
    8. SectionRetiring a dataset forty dashboards depend on7 minA migration story with real consumers. Staff loops ask for one.
    9. QuestionsData Pipeline · Failure & recovery15 minTell each incident as cause, fix, and the process you changed.

  2. Week 02 · ~3h

    Platform architecture: build, buy, or neither

    You can choose a platform shape with cost maths and named consumers.

    Skip this if you defend build versus buy with numbers, not preference.

    1. SectionFive questions that narrow seventy-two options7 minA method for tool choice that survives a sceptical panel.
    2. InterviewBuy the connector or write it6 minBuild versus buy, priced over three years, not one sprint.
    3. InterviewWarehouse or lakehouse6 minThe platform decision staff loops ask by name.
    4. SectionFive reference stacks, costed end to end8 minWhole platforms with prices. Borrow the shape, not the vendor.
    5. InterviewBespoke pipelines or a framework6 minForty sources: when a framework pays for itself.
    6. InterviewOne orchestrator or several6 minThree teams, three schedulers, one dependency. Decide.
    7. SectionRunning more than one executor6 minPlatform shape for many teams on one scheduler.
    8. ExerciseOne tenant ate the cluster50 minin the studioOne tenant starves the rest. Design the isolation.
    9. InterviewOne team's DAG ate everything6 minMulti-tenant fairness, asked the way platform teams live it.
    10. ExerciseIt has to be there by six45 minin the studioA critical path many teams wait on. Protect it.
    11. InterviewAirflow, Glue or Step Functions6 minManaged or self-run, justified by team and cost.
    12. InterviewOn-demand or provisioned capacity6 minCapacity commitments for many teams: price both options.
    13. SectionThe best pipeline is often no pipeline6 minStaff judgement includes stopping a project. Say when.
    14. QuestionsData Pipeline · Trade-offs & tool choice15 minEvery choice with its cost, its owner and its exit.

  3. Week 03 · ~3h

    Modeling for an organisation, not a project

    You can set a modeling standard several teams can build against.

    Skip this if you can argue mesh, Kimball or Vault for a given org.

    1. SectionThe four principles of data mesh7 minDomain ownership, products, self-serve, federated governance.
    2. SectionAnti-patterns, and when to adopt7 minStaff answers say when mesh is the wrong answer.
    3. SectionChoosing an approach, and blending them7 minAn org-standards decision, not a textbook definition.
    4. InterviewKimball or Inmon6 minSay what would change your answer. That is the test.
    5. InterviewWhen Data Vault earns its cost6 minAuditability against complexity, for a named organisation.
    6. SectionConformed dimensions and the bus matrix8 minThe standard every team models against.
    7. InterviewConforming two marts6 minTwo teams, two product dimensions. Unify without breaking both.
    8. SectionSurvivorship and the golden record7 minWhich system wins, decided once, for everyone.
    9. InterviewTwo dimensions become one entity6 minAn acquisition merges customers. Plan the identity change.
    10. ExerciseCustomer 360 tradeoffs45 minin the studioIdentity across systems, owned by more than one team.
    11. SectionMart, metric, or both?6 minOne number on every dashboard: the metrics-layer decision.
    12. ExerciseDatasets, owners, and access grants55 minin the studioOwnership and access modelled as data, across teams.
    13. InterviewHow do you review a model6 minYou will review models you did not write. Bring a rubric.
    14. QuestionsData Modeling · Approach trade-offs15 minArgue the fit for the organisation, not the textbook.

  4. Week 04 · ~5h

    Contracts, lineage and compliance across teams

    You can make other teams’ changes safe, visible and compliant.

    Skip this if you enforce contracts across producers without a central bottleneck.

    1. SectionThe producer-consumer handshake7 minA contract is an agreement between teams, then a schema.
    2. InterviewWho owns the data contract6 minTwelve models break overnight. Whose fault, and whose fix?
    3. InterviewShould the producer validate?6 minShift validation left, and say what it costs producers.
    4. InterviewWhat every event should carry6 minThe envelope every producing team must agree on.
    5. SectionEnforcement, lineage and governance7 minEnforcement that scales without you in every review.
    6. InterviewWho breaks if this column changes6 minLineage answers blast radius, before the rename ships.
    7. SectionDraw the split7 minWhere one project becomes several, and who owns each.
    8. InterviewOne project or several?6 minFour teams, 900 models, one repo. Decide the split.
    9. ExerciseThe payload changed overnight45 minin the studioThe payload changed overnight. Quarantine, then renegotiate.
    10. ExerciseIs this change safe?45 minin the studioThe gate every producing team passes before merging.
    11. SectionDeleting one person from everywhere7 minErasure is a lineage problem wearing a legal hat.
    12. InterviewGDPR erasure across seven years6 minAppend-only history meets a legal deletion. Design it.
    13. ExerciseDelete me from everywhere50 minin the studioDelete one person from everywhere, provably.
    14. InterviewWhere PII lives in the model6 minDecide where sensitive columns live before anyone queries them.
    15. ExerciseRisk and compliance audit50 minin the studioAn auditor asks what was true. Answer from the model.
    16. ExerciseDeletion that doesn't break the thread45 minin the studioDeletion that keeps history honest.
    17. QuestionsData Pipeline · Architecture design15 minName the owner, the contract and the blast radius.

  5. Week 05 · ~5h

    Migrations: getting fifteen teams to move

    You can migrate a shared table with evidence, a rollback and a sunset date.

    Skip this if you have led a migration many consumers depended on.

    1. InterviewExpand, migrate, contract6 minThe pattern behind every migration without a flag day.
    2. SectionHow to ship a breaking change anyway7 minExpand, migrate, contract, applied to a live schema.
    3. InterviewProving a migration was safe6 minEvidence before cutover, not confidence.
    4. InterviewRolling a migration back6 minWrong numbers at cutover. The rollback you planned earlier.
    5. ExerciseMoving without a gap60 minin the studioCut over without a gap, then prove it reconciled.
    6. Sectionv2 ships, v1 is pinned, nobody stops7 minVersioned models, and the consumers who never upgrade.
    7. InterviewVersioned models6 minDrop a column two teams read, with a sunset date.
    8. InterviewDeprecating a metric definition6 minTwo definitions in use. Retire one without a revolt.
    9. InterviewThe source deleted a row6 minAn upstream deletion policy, felt by every consumer.
    10. ExerciseChanging a pipeline that’s already running45 minin the studioChange a running pipeline: diff, deploy, roll back.
    11. ExerciseThe table everyone depends on45 minin the studioDeprecate a table other teams read, without an outage.
    12. SectionGetting from Airflow 2 to Airflow 37 minA platform upgrade under every team at once.
    13. InterviewPlanning the 2 to 3 upgrade6 minFour hundred DAGs. Sequence it, and handle stragglers.
    14. SectionNinety days, in the wrong order7 minRollout order matters. See what breaks when it is wrong.
    15. ExerciseYou fixed it. Who already used it?50 minin the studioRestatement: find every consumer of the wrong numbers.
    16. QuestionsData Modeling · Evolution & migration15 minEvery migration: evidence, rollback, sunset, owner.

  6. Week 06 · ~3h

    Cost and reliability at organisation scale

    You can attribute cost to teams and plan for ten times the load.

    Skip this if you have designed chargeback and a disaster-recovery budget.

    1. SectionThe six shapes a data bill takes7 minCredits, DBUs, bytes scanned: know how each bill grows.
    2. InterviewCharging the cost back6 minOne bill, six teams. Attribution changes behaviour.
    3. InterviewThe bill nobody attributes6 minCI and dev environments: the cost nobody owns.
    4. ExerciseIt works, and it costs too much45 minin the studioCut the platform bill without cutting what teams rely on.
    5. InterviewWhat saturates first at ten times6 minName the first bottleneck at ten times the volume.
    6. InterviewCapacity is not just throughput6 minHot keys cap you before throughput does.
    7. InterviewRetention as a cost decision6 minCompliance sets the floor. Tiering sets the bill.
    8. InterviewWhat a second region costs6 minDisaster recovery, priced before it is promised.
    9. SectionLineage: cause and blast radius7 minReliability at scale starts with knowing who is downstream.
    10. SectionSLIs, SLOs, SLAs and alerts people answer7 minPlatform SLOs other teams can plan against.
    11. InterviewWhat Airflow costs6 minThe bill doubled and the DAG count did not.
    12. ExerciseA ledger a regulator will accept45 minin the studioCorrectness a regulator accepts, at platform scale.
    13. QuestionsData Pipeline · Cost & scale estimation15 minEvery platform decision ends with its cost and its owner.

  7. Week 07 · ~3h

    The brief: present a project, then defend it

    You can present a project you led and defend its trade-offs.

    Skip this if you have presented a past project to staff engineers recently.

    1. ChapterStripe20 minStaff candidates write a brief, then present and defend it.
    2. ChapterSnowflake20 minA past-project presentation, then questions on trade-offs.
    3. SectionDesign lab: a whole design, end to end8 minA whole design, clarifying question to trade-off.
    4. SectionNarrating a design6 minNarrate for a panel that will challenge every choice.
    5. SectionAgreeing the handoff before you build6 minAlignment before building: the staff habit panels probe.
    6. ChapterLinkedIn20 minA craftsmanship round: rigour and quality culture, not only delivery.
    7. ChapterMeta20 minStaff behavioral probes leadership across teams.
    8. SectionWhat you now owe the next engineer6 minStandards that outlive you. Say what you left behind.
    9. ExerciseWhen you cannot do it all60 minin the studioDecide what not to do, and defend it to a panel.
    10. ChapterFollow-Up & Negotiation20 minStaff offers hinge on level. Negotiate the level first.
    11. QuestionsData Modeling · Critique an existing model15 minCritique like a reviewer: what breaks, who is affected, what it costs.

03

Start it now

Reading is free and needs no account. The plan itself tracks on your dashboard.

Start Staff Data Engineer · 8+ yearsTracked, 7 weeks, ~24hSee the other plans16 more, by what you are aiming at