Lakehouse Prepchief architect

#Self-Assessment and Gap Triage

Run this twice: at the start of the sprint and after the mock loop. The delta tells you whether the sprint worked; the absolute scores tell you what to do with your last day.


#The scorecard

Score 1–5. Be harsh. A generous self-assessment costs you the job; a harsh one costs you an afternoon.

#Platform depth

Topic Can explain at L1 (exec) L2 (director) L3 (engineer) L4 (internals) Score
Spark execution model
Catalyst / AQE
Joins and skew
Memory, spill, OOM
Photon
Delta transaction log / ACID
Data layout (clustering, partitioning)
Deletion vectors / MERGE
Structured Streaming
Unity Catalog model
Fine-grained access / ABAC
Workspace / metastore topology
Ingestion and declarative pipelines
Security and networking
GenAI / agent governance
Competitive positioning
Cost model and FinOps
2026 product currency

Rule: anything scoring below 3 at L3 is a gap that will show under pushback. Anything below 3 at L1 is a gap that will show with an executive. Both matter; the L1 gaps are faster to fix.

#Interview performance

Competency Score Evidence from the mock
Discovery before designing
Naming trade-offs explicitly
Depth on demand
Diagnostic method (evidence before fix)
Altitude range (exec ↔ internals)
Handling interruption
Handling being wrong
Objection handling
STAR stories with numbers
Time management
Questions asked of them

#Triage rules

With limited time left, fix in this order:

  1. Bluffing. Anywhere in the mock you pretended to know something. This is the only disqualifying category — fix it by rehearsing the "I don't know, here's how I'd reason about it" move until it is comfortable rather than shameful.
  2. Discovery. If you designed before asking in any scenario, drill five scenario openings — just the first three minutes, nothing else.
  3. The presentation. In a field loop this carries the most weight. Below 4 means rehearse, not read.
  4. The hands-on diagnostic. Your JD names it explicitly. Below 4, re-drill the six scenarios in ../01-platform/06-performance-and-cost.md §5.
  5. Numbers in STAR stories. Fast to fix, disproportionately noticed.
  6. L1 translations. Also fast — one pass through the translation table in ../02-drills/06-executive-business-case.md §1.
  7. Everything else. Knowledge gaps at L4 on topics you'll rarely be pushed to. Lowest priority, despite feeling like the most urgent.

Note the inversion: the things that feel most urgent (obscure technical gaps) are last, and the things that feel least like studying (rehearsing discovery, practising "I don't know") are first. That is not an accident — it is where the marginal scoring is.


#Readiness gate

You are ready when all of these are true:

  • You can draw all ten Tier-1 diagrams in four minutes each, from memory.
  • You can run any of the six diagnostic scenarios aloud without notes.
  • Your presentation lands in 30 minutes with six interruptions.
  • You have six STAR-L stories, each three minutes, each with a number.
  • You can answer "when would you not use Databricks?" with three genuine cases.
  • You can say what changed on the platform in the last six months and why it matters architecturally — without reciting a launch list.
  • You have five questions to ask them that show you understand the job.
  • You have said "I don't know" out loud, in practice, without flinching.

If six of eight are true, you are in good shape. Interviews are not passed by knowing everything; they are passed by being obviously credible and obviously honest. Chasing the last two boxes on the final evening usually costs more in sleep than it returns in performance.