Designing structure into an AI-led job application platform for candidates
Product Strategy
AI NAtive UX
I came in to rethink the visual design, fix the flows, and build the product's wider surfaces from scratch as it grew from a single tool into a workflow.
What is Resumatch.io?
Resumatch is an end-to-end application tool and job board. Candidates generate resumes tailored to specific roles, build shareable credential-validated profiles, and manage their applications in one place. It is built to carry someone through an application, not just to produce a document.
The Brief
I worked directly with the founder and lead engineer in a tight agile loop to redesign existing flows and add new features. The tech team was AI-driven, so builds and iterations shipped fast, often in response to real-time user behavior. That speed is an advantage and a risk. Ideas arrive faster than they can be structured, and a product can accumulate features faster than it accumulates logic.
Most of what I did was convert ambiguity into structure. The founder brought direction and a steady stream of feature ideas. My job was to give them a user point of view and a shape that could ship without the product fragmenting.
Not every problem got solved simply because it existed. Ideas were prioritized and some were deliberately held back. Because I worked in direct tandem with the founders and engineering, design decisions were made against real technical constraints.
Building a system from a surface
V1 worked but carried broken flows and a limited visual system, with the heaviest overload in the most feature-dense areas. Before adding anything, I set foundations: a typographic scale, a color system, iconography, and consistent spacing.
In the dense editing surfaces, the priority was subtraction, reducing competing elements so the primary action in any view was unambiguous. We gave every later feature in an increasingly complex idea a place to land instead of being bolted on.
Considering the key factor- trusting AI with your Resume
A resume is the key first step in any job search and a differentiator in the current pools of generic content.
When Resumatch tailors content to a job description, the risk isn't bad writing. It's fabricated or generic claims that are repeated across 100s of other documents.
I designed the AI editing layer around building confidence and trust:
Suggestions appear inline, in context. The original is struck through and the AI's version sits directly beside it, so the user always sees what changed, not just the result.
Control is granular. Each suggestion can be accepted or discarded on its own, or handled in bulk with accept-all, discard-all, or "insert with original" for users who want to keep both versions and decide later.
A live preview renders the real document. As edits are accepted, the formatted resume updates alongside, with template, size, and font controls. The abstract review never loses its connection to the final output.
The uMatch score sits next to the preview, tying every edit back to the only question that matters: does this move me closer to the role.
The Application Kit: a new way of standing out and proving credibility and value
We built the Application Kit to turn each candidate profile into a public landing page built for credibility: validated credentials, AI interviews and assessments that demonstrate skill, and the uMatch score shown against a given role.
Candidates can build a kit for a specific job by uploading the JD, or a general kit for broad outreach. Privacy settings control visibility, and viewers with access can refer the candidate directly from the page.
The design challenge was making this read as evidence, not as a gimmick. The page had to feel like proof of a person, not a marketing landing page for one.
Baking the new features into the core structure of the app
When I started, the homepage was a single action: upload a resume or a JD. That was correct for a product that did one thing- build personalised and job-specifc resumes.
As the Application Kit, assessments, and matching arrived, the product's center of gravity shifted. It stopped being a tool you used once and became something you returned to across the life of your job search, and eventually, beyond. The home had to hold an in-progress job search, so it evolved into a tracker.
That evolution reflects how the whole product was built. Nothing was designed 0-to-1 with a fully specified vision. Features were introduced progressively and folded into the existing architecture as they arrived, rather than forcing a rebuild each time. The work was making sure every addition had a coherent place to live.
The tracker is manual today, with automation in progress, but the structural decision was the point: recognizing the moment a tool becomes a workflow, and restructuring the product's front door to match.
Outcome
Resumatch shipped and scaled, and it is still actively evolving. The product reached 8,000+ users and now generates roughly 40,000 resumes per day. User feedback has centered on exactly what the work was aimed at, the interface-
"
The experience feels simple without feeling thin. It helped people get to a better resume faster, with fewer moments of hesitation.
"
Extremely easy to use.
"
Very easy to use, excellent review and modifications.
Closing Thoughts
Resumatch is what early-stage, AI-native products usually need and rarely get: someone to hold the user's point of view steady while everything else moves fast. The visual system kept the product coherent as it grew. The editing layer made a risky AI feature trustworthy. The Application Kit took a loose idea, a profile of some kind, and turned it into a defensible product surface. The tracker turned a set of separate tools into a single workflow.
The throughline was structure. Taking an ambiguous, fast-moving brief and giving it a shape that users could trust across an entire application, not just a single document.
