Skip to content
AI Engineer Guide.

concept · career · intermediate · concept 3 of 33

Resume & GitHub

0 of 33 complete

Why it matters

A repository link is often the only artefact a reviewer can inspect before an interview. They are not looking for volume; they are checking whether the code can be understood, run and trusted. A project with a clear README, a seed script and a test suite demonstrates more than a larger project with none. The profile is a sample of your defaults, which is why the presentation matters as much as the code.

Mental model

A profile is a shop window: the pinned repositories are the display and the README is the label. The metaphor breaks because a window is arranged once, while a profile accumulates whatever you last pushed, so neglect is visible and the display needs maintenance rather than a single arrangement.

How it works

Mechanism, step by stepInteractive
  1. 1Pin three repositories that represent different strengths, and archive or make private the ones that do not.
  2. 2Write a README with the problem, a quickstart that works from a clean clone, the results and the limitations.
  3. 3Add a seed script or fixture data so a reviewer can reproduce the result without your environment.
  4. 4Include tests for the core behaviour and a command that runs them, even if coverage is modest.
  5. 5Clean commit messages on the pinned projects so the history reads as a sequence of decisions.
  6. 6Add a short note on what you would do next, which shows judgement about scope rather than abandonment.

Step 1 of 6

Pin three repositories that represent different strengths, and archive or make private the ones that do not.

Step 1/6 · step 1

Figure. The mechanism in this page's own steps. The figure adds sequence, not new claims.

What to notice. A step can be read alone, but the order is the mechanism. The control lets you stop on any one.

Text version

The page's steps: 1. Pin three repositories that represent different strengths, and archive or make private th… 2. Write a README with the problem, a quickstart that works from a clean clone, the results … 3. Add a seed script or fixture data so a reviewer can reproduce the result without your env… 4. Include tests for the core behaviour and a command that runs them, even if coverage is mo… 5. Clean commit messages on the pinned projects so the history reads as a sequence of decisi… 6. Add a short note on what you would do next, which shows judgement about scope rather than…

  1. Pin three repositories that represent different strengths, and archive or make private the ones that do not.
  2. Write a README with the problem, a quickstart that works from a clean clone, the results and the limitations.
  3. Add a seed script or fixture data so a reviewer can reproduce the result without your environment.
  4. Include tests for the core behaviour and a command that runs them, even if coverage is modest.
  5. Clean commit messages on the pinned projects so the history reads as a sequence of decisions.
  6. Add a short note on what you would do next, which shows judgement about scope rather than abandonment.

Trade-offs

ChoiceGood forCostFails when
Few polished repositoriesClear signal and easy reviewLooks less prolificThe role screens for open-source volume
Many small projectsShows range and curiosityDilutes the signal and invites low-quality readsA reviewer only opens the first one
Forked librariesShows you work in real codebasesAttribution and maintenance questionsThe fork has no visible contribution
Private work described in proseProtects employer IPUnverifiable claimsThe role requires code samples
Trade-off mapInteractive

Good for

Clear signal and easy review

Cost

Looks less prolific

Fails when

The role screens for open-source volume

  1. The role screens for open-source volume
  2. A reviewer only opens the first one
  3. The fork has no visible contribution
  4. The role requires code samples

Figure. Each choice with what it buys, what it costs, and the case where it loses.

What to notice. The fourth column is the decision: if the losing case describes your situation, the choice is wrong.

Text version

Four trade-offs: Few polished repositories is good for Clear signal and easy review, costs Looks less prolific, and fails when The role screens for open-source volume. Many small projects is good for Shows range and curiosity, costs Dilutes the signal and invites low-quality reads, and fails when A reviewer only opens the first one. Forked libraries is good for Shows you work in real codebases, costs Attribution and maintenance questions, and fails when The fork has no visible contribution. Private work described in prose is good for Protects employer IP, costs Unverifiable claims, and fails when The role requires code samples

Failure modes

  • Broken quickstart. The README instructs a command that fails on a clean machine. Debug: clone into a fresh directory and follow the README exactly.
  • No entry point. The repository has files but no statement of what it does. Debug: read the first paragraph and check whether it names the problem and the user.
  • Stale pinned repos. The visible projects are years old and unrelated to the role. Debug: compare pinned repositories with the technologies named on the resume.

In production

Treat the profile as a release artefact: re-check the quickstart quarterly and after dependency upgrades. Keep a CI badge only if the pipeline is actually green.

Common misconceptions

People think: A green commit graph is a strong signal.

Actually: Reviewers read content and clarity; a dense graph of trivial commits says little about judgement.

People think: Open source is required for AI roles.

Actually: A small, well-documented project with an evaluation is sufficient evidence and often more legible.

Interview angle

Strong answers can walk through one repository: the problem, the design choice, the test and the limitation. Weak answers point at a profile and cannot explain a design decision inside it.

Sources