The resume, rewritten as git log
Organised by problem rather than by employer. Every line is a real repository with a real creation date, and the through line only becomes visible once they are laid out chronologically: the same modelling problem, three levels of abstraction apart.
Every repository line is real, with its creation date read from the GitHub API. The through line is visible once it is laid out chronologically: identifier utilities in 2023, a brand knowledge graph in 2024, and a full deployment and governance platform by 2026. Same problem, three levels of abstraction apart.
A resume flattens the thing that matters
Role, company, dates, three bullets. That format hides continuity, because it is organised by employer rather than by problem. The problem is what actually persisted.
Continuity is the signal
Four employers over four years reads as movement. The same modelling problem attacked from four vantage points reads as a specialisation, and it is the more accurate description.
Commits are falsifiable
A bullet point is a claim. A repository with a creation date and a history is evidence, and anyone can go check it.
Abandoned branches count
Several of these went nowhere. Leaving them in is more honest than a portfolio that only shows what worked, and the failures are usually where the method got sharper.