Employer-facing platform that connects recruiters with the university network

Historically, Link 1 operated as a manual, managed service. Because it was built on legacy tech and shared codebases, it was impossible to maintain UI consistency or build modern components. This created a fragmented experience where our internal product admins had to handle the posting process on behalf of every client.

Link 2 is the complete rebuild of that experience. My focus was moving the platform from that manual model into a modern, self-serve SaaS product, taking a complex, data-heavy workflow and making it intuitive enough for employers to sign up, post, pay, and track their own performance without needing a human safety net.

The main problems were:

• Low post quality: users often write poor, incomplete job descriptions. Admins used to fix these manually; now the interface has to prevent them from being submitted in the first place.
• University rejections: universities have strict criteria for what they show students. If self-serve posts don’t meet those standards, they get rejected, and high rejection rates would damage our university partnerships.
• Rigid tech stack: the legacy codebase was too stiff for modern UX. We couldn’t iterate or build the components a self-serve model needed.

I watched admins spend hours correcting typos and re-writing vague job descriptions so they wouldn’t be rejected by universities. Quality was the most valuable thing they did. If we were going self-serve, the UI had to catch those mistakes before the user posts the role.

Account managers had dozens of tabs open, copying and pasting snippets from old job descriptions and cross-referencing spreadsheets just to fill out one form. Even a simple post took at least twenty minutes, and missing information meant days chasing clients. Because the internal reporting tool was unreliable, the team was trapped in a cycle of manual exports and weekly calls just to explain basic performance.

How we turned tedious data entry into an engaging UX. The Link 1 interface forced users to process too much information at once. I moved to a multi-step architecture built on progressive disclosure, chunking the data into thematic stages to lower cognitive load and create momentum. The UI remembers recurring data like company benefits and office locations, so users only touch what’s unique to the new role. And for the blank-page problem, a user can drop a link or raw job spec into a single field and watch the system extract and map it into the form in real time, a deliberately transparent interaction that builds trust in the automation.

Live scoring. A floating sidebar shows live quality scoring as the role is built. At review, the AI gives a final score and pulls out the exact sections that need changing, rather than bouncing users back through the form at the point of payment or publish.

Role statuses and application statuses live on the dashboard and in job details, with payment status and contextual tips alongside. A simple ruling system pairs every state with a clear next action, so recruiters and admins always know whether a role is live, in review, or needs something from them, without a support ticket.

Identifying the problem: the university moderation flow was already taxing under the managed service, and self-serve volume would multiply it. I mapped the entire approval cycle end to end, then redesigned the flow so quality is enforced upstream. The coach, scoring, and compliance checks catch issues before a post ever reaches a moderator, and what remains for universities is a fast, criteria-led review rather than a rewrite.

I designed the new reporting suite on a Bento Box layout built for high-density information: high-level metrics glanceable, granular data one click away. Benchmarking charts use value anchoring, placing the user’s performance on an axis against their peers, giving instant context that used to take an account manager an hour to explain. The internal team now works from the same real-time suite as clients, which eliminated the spreadsheet trap entirely.