Build a portfolio evidence index for interviewers

A reviewer should not need to search an entire repository for the split, metric or reproduction command behind a claim. An evidence index turns the project landing page into a map from statement to artifact.

Require a location for every verified claim

The career evidence lab checks an invented index.

from career_cases import portfolio_index_case

result = portfolio_index_case()
assert result["verified_rows"] == 2
assert result["unsupported_verified_rows"] == 0
assert result["index_passes"] is True
print(result["rows"])

The chronological-split claim points to evidence/split-manifest.json; the test metric points to evidence/metrics.json. Review approval has no artifact and is explicitly not claimed. A blank location cannot carry “verified” status.

Index decisions, not every file

Useful rows include:

Link to stable relative paths. Record the command that regenerates each derived artifact and the environment it used. If an artifact is large or private, provide an allowed metadata record rather than copying sensitive material.

Make status explicit

Use states such as generated, test-passed, human-reviewed, approved or not claimed. Automated checks do not imply editorial or domain approval. Keep failed evidence, including a rejected model, with a clear status so the index does not become a catalogue of only favourable outputs.

Add a short “start here” path: read the decision brief, inspect the baseline and final evaluation, run the reproduction command, then review known limits. Test the path with someone unfamiliar with the project.

The Data Science course uses evidence manifests and model cards as part of capstone defence. The index makes those outputs accessible without replacing their technical detail.

Exercise

Index ten claims from one project. Break one path and remove one artifact, then run a link/status check that fails. Ask a peer to verify three claims without your help and record where the index was ambiguous.

Continue learning

This article is part of the Data science careers and portfolio decisions sequence. Use the neighbouring tasks when you need the prerequisite or the next application.

• Review the prerequisite or neighbouring task in Read a machine learning job description critically.
• Continue with Decide whether to specialize in NLP, forecasting or tabular ML.

Reference: Model Card Toolkit documentation.