Demonstrating capability.
WHAT DEMONSTRATES LITTLE
Tutorial reproductions Well-known teaching datasets Notebooks with no context
WHY
Everyone has those, and they show no judgement.
WHAT DEMONSTRATES CAPABILITY
A problem you framed yourself Messy data you obtained and cleaned Honest evaluation, including what failed A deployed, working system Documentation explaining the decisions
WHAT TO DEPLOY
Something small, reachable at a link.
WHY
It proves you can get past the point most projects die.
WHAT TO WRITE ALONGSIDE
Why you framed the problem as you did What you tried that did not work What the limitations are
WHY FAILURES SPECIFICALLY
They demonstrate judgement, which successes alone do not.
WHAT PROBLEMS TO CHOOSE
Ones from a domain you understand Ones using data you can legitimately obtain Ones small enough to complete
WHAT TO AVOID
Projects requiring data you cannot get Projects too large to finish
WHAT TO INCLUDE FOR LOCAL RELEVANCE
Problems from the local context, which are distinctive and less commonly attempted.
