Analyst

Data / Analytics + AI

Turn tables into decisions. Use AI for narrative, not as an unchecked oracle.

2 weeks · 12 lessons · 4 labs · query-then-prose project

0/12 lessons on this path

Mastery syllabus

1. 1 · Maps and measures

   Embeddings and evals keep the story honest.

   • Neighbours in meaning-space10 min
   • Proof over confidence11 min
   • One team brain10 min

2. 2 · The card that funds you

   Before/after on one card.

   • Hours you can show10 min
   • One painful weekly task9 min
   • Write proof on the resume10 min

3. 3 · Query first

   Numbers leave the warehouse. Empty means silence.

   • Numbers leave the warehouse first12 min
   • Lock the KPI before the sentence12 min
   • Ask for a form, not a poem12 min

4. 4 · Refuse a pretty lie

   A director can love a wrong chart.

   • Closed book vs open notes12 min
   • Where AI must not decide9 min
   • The world moved. Did your model?12 min

Simulation labs

• Similar is not same

  Nearby points still need a check.

  Open the lab

• Hours and error rate

  Two numbers beat a dashboard poem.

  Open the lab

• Warehouse first

  Query, then prose.

  Open the lab

• Lock the KPI

  Pretty is last.

  Open the lab

The project that hires

One workflow: question → warehouse → prose, with a refuse

1. Pick one recurring question the team already asks.
2. The model may only speak numbers that came from the query.
3. Show one refused answer.

Interview room

• When is SQL the right tool and the LLM a dangerous one?
• Design a check that a generated chart matches the warehouse.
• Walk query → lock KPI → prose → refuse.
• A denominator is missing. Ship or stop?

Resume lines that hire

• Used AI to draft insights, then verified every figure against source SQL.