#The Ten-Day Sprint

Assumes ~3 hours on weekdays and ~5 on weekend days. If you have fewer, cut from Deepen first, never from Drill — reading feels productive and is the least productive thing here.

The governing ratio: 40% input, 60% output. If you are reading more than you are speaking and drawing, you are preparing to recognise answers, not produce them.

#Daily shape

Record yourself for every Drill block from Day 3 onward. Phone voice memo is enough. You will hate it and it is the single highest-yield thing in this plan — you cannot hear your own filler words, your pace, or the moment you lost the thread without a recording.

#Day 1 — Orientation and honest baseline

• Deepen: 01-role-and-competency-map.md and 02-interview-loop-playbook.md.
• Do: fill in the competency grid and the evidence inventory. Non-negotiable — everything later depends on having your own examples at hand.
• Drill: record your 90-second opening pitch. Then re-record until it is 90 seconds, not 150.
• Close: read ../01-platform/09-platform-2026-whats-new.md once. You are clearing stale vocabulary early so you stop using old names all week.

Baseline honesty: if the competency grid has more 2s than 4s in rows 1–3, shift an hour a day from Drill to Deepen for Days 2–4, then switch back. Do not skip the drills entirely.

#Day 2 — Spark internals

• Deepen: ../01-platform/01-spark-internals.md.
• Drill: explain aloud, 90 seconds each, no notes: Catalyst's four phases · what AQE changes at runtime and why it needs shuffle statistics · why a broadcast join fails · what Photon does and when it does not help · what a shuffle physically is.
• Close: draw the query lifecycle from SQL text to task execution, from memory.

#Day 3 — Delta Lake internals

• Deepen: ../01-platform/02-delta-lake-internals.md.
• Drill: the classic sequence, spoken end to end: "What's in _delta_log?" → "Two concurrent writers?" → "Why is time travel bounded?" → "Z-order vs liquid clustering?" → "What do deletion vectors change about MERGE?"
• Close: flashcards on protocol/table features, retention defaults, isolation levels.

#Day 4 — Performance and debugging (the flagship day)

• Deepen: ../01-platform/06-performance-and-cost.md — all of it.
• Drill: work all six diagnostic scenarios in that file out loud, using the Symptom→Scope→Evidence→Hypothesis→Test→Fix→Prevent frame. Time-box each to 6 minutes.
• Close: memorise the triage order (§2). This is the material most likely to decide a "remain hands-on" role.

If a day slips, do not let it be this one. Move something else.

#Day 5 — Lakehouse topology, Unity Catalog, security

• Deepen: ../01-platform/03-lakehouse-architecture-and-topology.md
  • ../01-platform/04-unity-catalog-governance.md
  • skim ../01-platform/08-security-networking-compliance.md.
• Drill: draw, from memory, in under 5 minutes each: control plane vs compute plane · a three-environment multi-workspace topology with UC · a network-isolated regulated deployment.
• Close: be able to answer "how many metastores do I need?" with the reasoning, not a number.

#Day 6 — Ingestion, pipelines, and the AI stack

• Deepen: ../01-platform/05-lakeflow-ingestion-and-pipelines.md
  • ../01-platform/07-ai-and-agents.md.
• Drill: design aloud — a CDC pipeline from an operational Postgres to governed gold tables with a 5-minute freshness SLO; then a RAG system over 2M internal documents with access control that respects existing permissions.
• Close: flashcards on streaming semantics and on GenAI governance.

#Day 7 — Architecture scenarios (first full drill day)

• Deepen: skim ../01-platform/10-competitive-positioning.md.
• Drill: six scenarios from ../02-drills/01-architecture-design-scenarios.md, 25 minutes each, spoken and drawn, model answer unread until you have finished. Then compare and write down only the gaps.
• Close: list your three weakest scenario types; those get revisited on Day 9.

#Day 8 — Build the presentation

• Deepen: ../02-drills/05-the-panel-presentation.md.
• Build: the deck. 10–14 slides, the structure given in that file. Do not exceed 14.
• Drill: deliver it once to a wall, timed. It will be too long. Cut 30%.
• Close: write the five questions you will open with in discovery.

#Day 9 — Objections, behaviourals, and weak spots

• Deepen: ../02-drills/03-customer-objections.md
  • ../02-drills/04-behavioral-star-bank.md.
• Drill A: have someone fire 15 objections at you at random. No preamble, 60-second answers.
• Drill B: record six STAR-L stories, one per value. Three minutes each, hard stop.
• Drill C: the three weak scenario types from Day 7.

#Day 10 — Full mock and taper

• Morning: the complete timed loop in ../03-practice/01-mock-loop-full.md. Score yourself against the rubrics. Get a human to play the panel if at all possible — pay someone if you have to; it is worth more than another day of reading.
• Afternoon: fix only the top three gaps. Resist the urge to re-read everything.
• Evening: stop. Read ../99-reference/cheatsheet.md once. Sleep.

#If you have five days, not ten

Merge as follows and accept the narrower base:

#If you have two days

1. Cheatsheet → performance playbook → presentation build.
2. Rehearse presentation twice → three scenarios spoken → 20 objections → sleep.

#The day before

• No new material. Nothing. New facts the night before displace rehearsed ones.
• Re-read the cheatsheet and your evidence inventory.
• Say your opening pitch and your six STAR stories out loud once each.
• Check the logistics: link, time zone, who is in the room, deck format and a PDF fallback.
• Prepare your questions for them.

#The hour before

• Cheatsheet, one pass.
• Your three strongest numbers, said aloud.
• One deep breath and this frame: you are a consultant meeting a client, not a candidate being examined. The room reads that posture, and it is also simply the job.