Prove
LiveProof-of-Work · Oct–Nov: every verified proof counts. 35 days leftAll outcomes
Skills

Ship Your Data Science Final Project

12 weeks · 3 milestones

Answer a real question with data, share the answer with the people who need it, then pass a blind review. The last step of the Varsity Curriculum: Data Science.

Milestone map

Milestone map

3 milestones

Pick a real question and plan it

1–2 weeks (10 hrs/week)

Pick a question real people need answered that your data skills can answer. It can be an analysis, a model or a data tool. Write a one-page plan: the question, who needs the answer, the data you will use and how you will know you got it right.

Proof required

Link your one-page plan, the data source, and notes from talking to at least two people who need the answer.

What gets checked

  • The question is real and the people are named (first names are fine).
  • The data exists and you are allowed to use it.
  • Success is something you can measure.

Common mistakes

  • A question nobody asked.
  • Data you cannot get or may not use.

Powstik Guide

A one-page plan for your final project.

Steps

  1. List three questions you or people near you want answered.
  2. Talk to two people who need the best one.
  3. Find the data and check you may use it.
  4. Write the plan on one page.
  5. Ask three people to check it.

Template

Question:
Who needs the answer:
Data I will use (and its licence):
How I will know I got it right:
What I will not do:

What gets sent back

  • No real people need it.
  • No data, or data you may not use.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

  • Is the question real?
  • Is the data real and usable?
  • Is success measurable?

You'll sign in first, then come straight back here.

Answer it and share it

6–10 weeks (15–20 hrs/week)

Do the work: get the data, clean it, analyse or model it, and share the answer with the people who need it. Keep everything in a public repo that reruns from start to end.

Proof required

Link the public repo that reruns the whole analysis, the shared result (report, dashboard or app), and proof that at least three of the people who needed it used it.

What gets checked

  • The repo reruns from raw data to the result.
  • Results are checked on data the model or analysis did not see.
  • Feedback from the people led to at least one change.

Common mistakes

  • Months of modelling with no one using the answer.
  • Results that only work on one laptop.

Powstik Guide

A public, rerunnable project whose answer three real people used.

Steps

  1. Share the simplest useful answer first.
  2. Get three people to use it.
  3. Fix the top problem they hit.
  4. Make the repo rerun from raw data.
  5. Ask three people to check it.

Template

README
Question:
Data (source and licence):
How to rerun it:
Result:
Who used it and what changed:

What gets sent back

  • No real users.
  • Cannot be rerun.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

  • Does the repo rerun?
  • Do the numbers hold up on held-out data?
  • Is the user proof real?
  • Is it the person's own work?

You'll sign in first, then come straight back here.

Defend it in a blind review

1–2 weeks (10 hrs/week)

Write a short report: the question, the data, what you did, what you found and what you would do next. Reviewers who do not know you read it and rerun the project.

Proof required

Link your final report (2–4 pages) and a 5-minute video walking through the project and its code.

What gets checked

  • The report matches the project and the numbers are real.
  • The video shows the code and the data, not just the charts.
  • Limits and next steps are honest.

Common mistakes

  • Claims the numbers do not back up.
  • A video that hides the code.

Powstik Guide

A final report and a 5-minute code walk-through.

Steps

  1. Write the report from your plan and results.
  2. Record the walk-through.
  3. Have a friend read the report cold.
  4. Submit for blind review.

Template

Final report
Question:
Data:
What I did:
What I found:
What went wrong:
What I would do next:

What gets sent back

  • Numbers without sources.
  • No code shown.

Resources

Foundationstart here

What a verifier looks for

  • Do the numbers in the report hold up when you rerun it?
  • Does the code match the video?
  • Is the person clearly the builder?
  • Would you trust this person with the next data question?

You'll sign in first, then come straight back here.