Loubby's job application flow required manual resume uploads — a friction point that dropped candidates before they ever applied. The team needed AI-generated resumes from plain descriptions, a live editing experience, and a full onboarding flow for a new agent marketplace. Both had to ship within the same delivery window.
Candidates on Loubby were abandoning the application flow at the resume step. Manual uploads required a polished, ready document — a barrier that cut off a large portion of users who had the skills but not the CV. The business needed AI to close that gap: a user describes what they do, AI produces a professionally structured resume they can edit, download, or submit immediately.
Simultaneously, the company was launching an agent marketplace and needed a full onboarding flow — separate logic and screens for buyers and sellers — designed and built from scratch alongside the resume work.
Rather than generating a resume and handing back a static document, I built the output directly into a live editing surface. The user sees their resume take shape in real time and can refine it immediately — no download, re-upload, or switching tools. This meant the AI layer and the editor had to be designed together from the start, but the result is a seamless experience: describe yourself, watch the resume build, edit what you want, apply. That tight loop is what turns a feature into something users actually complete.
A complete user journey over a technically simpler feature drop.
I would have set a hard file size rule from the start. The resume editor grew fast — generation, editing state, highlight markers, template logic all lived closer together than they should have. By the time files hit 900–1500 lines, refactoring was no longer optional and it ate into feature time. A 250-line limit enforced early would have forced the right separations before they became painful.