Product case study · Working web app

Résumé Forge

A résumé and application workspace built around one reusable source of truth—and controls designed to keep AI-assisted tailoring honest.

Open live app ↗View repository ↗

01 · The problem

Tailoring should not create a new truth every time.

Typical résumé workflows produce near-duplicate files and make it easy for generated language to drift beyond the underlying experience. The product starts from a reusable accomplishment bank, then treats every tailored document as a controlled selection from that record.

02 · Decisions I owned

Defined the core product model: one reusable accomplishment bank instead of a folder of drifting résumé copies.
Specified honesty controls, including adjustable conservatism, editable recommendations, per-bullet decisions, original-text recovery, and undo.
Designed the application-tracking workflow and decided which features shipped across successive versions.
Specified evidence-grounded interview preparation, including when it runs, how deep it goes, which evidence it uses, and how users recover from a stalled request.
Tested releases, found résumé-voice and parsing failures, and required corrections before treating a version as ready.

03 · What exists

Imports résumés from PDF, DOCX, images, and text.

Supports structured editing, section reordering, and job-description tailoring.

Generates editable cover letters and stores application materials by job.

Generates likely interview questions and evidence-grounded answer outlines from each saved application.

Exports PDF and DOCX files and supports private, account-based persistence.

04 · Product evidence

Résumé Forge showing job-tailoring recommendations beside the generated résumé preview — Full auto-tailor returns visible recommendations beside the résumé preview, leaving the user in control of what gets applied.

Résumé Forge showing evidence-grounded interview preparation for a saved application — A saved application can produce likely interview questions and answer outlines tied back to evidence from the submitted résumé and master CV.

05 · What I learned

Product ownership is not the same as code authorship. The useful skill here was turning a fuzzy problem into explicit behavior, testing whether the implementation matched it, and refusing output that sounded polished but was not true.

06 · Current edge

The next phase is deeper technical ownership: tracing the architecture end to end, documenting the main data flows, and re-testing sign-in, save, tailoring, and export before broader promotion.

Next case study

Process Notebook →