What I did

AI-Assisted Workflow

I generate early wireframes and design system output directly from the BA's brief, then bring those into client calls to work through page-by-page. The AI gets me to a starting point fast — the real work is deciding what to keep, what to challenge, and what to rebuild once the conversation reveals what the brief didn't say.

Stakeholder Alignment

Four departments meant four sets of priorities competing for the same pages. I worked through where their needs conflicted, sequenced what could share a pattern versus what genuinely needed its own, and kept the brief's gaps from turning into four different interpretations of the same site.

Design Judgment

Not every AI-generated screen or client request held up. I pushed back on a banner sizing request that would've undercut the page's hierarchy, and caught places where AI-generated structure didn't match real technical or content constraints — correcting those before they reached development.

Astra is a website for a B2B agricultural equipment dealer covering four distinct business lines — new equipment, used equipment, parts, and service — each run like its own department with its own priorities. I'm designing it from an intentionally incomplete brief: the business analyst fills in page details as she writes them, so waiting for a finished spec wasn't an option. I use AI to move from that partial brief to a rough structure fast enough to start a real conversation, then correct course based on what that conversation reveals.

Astra sells new equipment, used equipment, parts, and service — four business lines that operate almost like four separate companies sharing one site, each with its own priorities and its own idea of what the homepage should prioritize.

The brief I started from wasn't a finished spec. The business analyst had mapped the general structure, but across the document, entire sections were marked "as per design" or "as per agreement with designer" — decisions she'd deliberately left open because she couldn't resolve them without input from other units, or without a visual to react to first. In practice, that meant my first job wasn't to deliver a prototype. It was to produce something concrete enough to start the conversations that would actually fill those gaps — a visual the client, the BA, and the other departments could look at, disagree with, and react to, so the real requirements could surface at all.

"As Per Design"

Part One — SAT

The first audience: recruiters juggling candidates across multiple of client companies.

I ran a two-hour workshop with 18 of Ework's recruiters, opening with a quick icebreaker (a round of memes, just to loosen people up) before asking everyone to walk me through what actually happens when a candidate applies — their real day-to-day process, in their own words. From there, I laid out a set of potential problems — missing follow-ups, forgetting which client a candidate belonged to, double-booking interviews, losing track of candidates between companies — and had the group dot-vote on FigJam to surface what actually hurt the most. Losing track of candidates between companies and forgetting which client they belonged to came out clearly on top. I reframed those into "How might we" statements to keep the group focused on the problem before jumping to solutions, then opened it up to free discussion.

The recurring ask was simple: an easy way to switch between companies and between job postings within a single company — several people described wanting something like tabs or a switcher. We closed with each person naming the one thing they most wanted from the new system, plus a quick retro on the session itself. The scale of the problem became clear from the numbers that came out of it: recruiters were juggling requests from 2–3 client companies at once, and losing 1–3 hours a day to manual tracking just to stay on top of it.

Recruiter Research

Option A — Everything in one screen

Concept Testing

This version followed what recruiters described almost literally: company and job posting switching lived in persistent tabs and chips at the top, with the full candidate list and a status filter always visible below. It matched what people asked for in interviews — nothing hidden, no extra clicks to see where you were. The tradeoff was screen real estate: as soon as a recruiter was juggling more than 3–4 companies, that tab row would need to scroll or wrap, and the "everything visible" promise started to compete with the actual candidate data underneath it.

Option B — Left-hand navigation, grouped drill-down

This was my alternative to what recruiters had literally asked for: a collapsible sidebar listing companies, each expandable to its open job postings, so a recruiter picks a company → job before ever seeing candidates. It cost an extra step compared to Option A, but it kept the workspace uncluttered and scaled cleanly regardless of how many client companies someone was managing — the exact ceiling Option A ran into.

Testing both head-to-head, 70% of recruiters preferred Option B — the structure I'd proposed rather than the one they'd described in interviews. It's a useful reminder that what people ask for in research is a description of their pain, not necessarily the right shape for the fix; the drill-down solved the same "where am I" anxiety without the persistent-tab tradeoff, once they actually used it instead of just talking about it.

Clicking a candidate card opens a full-screen view with everything a recruiter needs to make a decision in one place — the same fit-comparison table from the card, plus cover letter, CV, portfolio link, and the candidate's answers to the application questions, with a match-percentage stepper up top so the overall score isn't lost among the details.

For the document links, I deliberately chose "open in a new tab" over a direct download. Recruiters aren't archivists — they need to review a CV once, decide, and move on, not accumulate dozens of downloaded files cluttering their computer. Opening in a new tab lets them review and download only the ones they actually want to keep. The call-to-action is a single, prominent email button rather than a fuller contact card: when I asked recruiters in interviews how they actually communicate with clients, the answer was consistently "directly through email" — so the button opens straight into their default mail client rather than a chat widget or in-app messaging system they don't use for this.

Candidate Detail View

Part Two — Verama

The same system needed to work for a second audience — not recruiters managing candidates, but the consultants themselves.

I ran a mixed-methods survey — open-ended and closed-ended questions — with 200 freelancers already using the platform, to understand where the experience was breaking down.

Research

• Payment timing was confusing. Freelancers worked one week and got paid partway through the following week — but the wording around this schedule wasn't clear, so people frequently didn't understand when money was actually coming.

• Job applications disappeared into a black hole. Freelancers applied to jobs and often never heard back — with no way to track what they'd applied to, where each application stood, or who had gone quiet. People specifically asked for a visual way to see their application pipeline: applied, interviewed, technical round, rejected, ghosted — something no competing platform offered at the time.

• Freelancers wanted community, not just a marketplace. Many were also looking for standard full-time work, not just gigs — and wanted a space to exchange experience with peers. This led to Verama+, a hybrid between a blog and a forum: helpful articles alongside a space to connect and share experience

Three findings stood out:

Verama ran on the same design system as SAT, extended and applied as the product evolved — I worked alongside other designers to grow the system while adapting it to Verama's specific needs, rather than treating it as a separate build from scratch.

The design cycle followed a consistent path: prototype, present the idea, defend the reasoning to leadership, then move to high-fidelity design once direction was approved — followed by development handoff.

One decision worth calling out: my first proposal for application tracking was a Kanban-style board — cards moving automatically through Applied, Interview, Technical Interview, Hired, or aging into Ghosted if a company went silent. Leadership pushed back — the automatic status transitions were a much bigger engineering lift than the visual itself. Rather than dropping the idea, we scoped it down: ship a simpler card list first, giving freelancers a way to see what they'd applied to without losing track — the real pain point — and treat the full automated board as a future iteration once the underlying tracking logic existed to support it.

Design & Delivery

SAT replaced spreadsheets, email threads, and an unused legacy tool for recruiters juggling 2–3 clients at once and losing 2–3 hours a day to manual tracking. Testing two navigation concepts head-to-head showed 70% preferred the structured drill-down over what recruiters had described in interviews — proof the fix beat the literal ask.

Verama surfaced three pain points through a 200-freelancer survey: confusing payment timing, no way to track applications, and demand for community beyond the marketplace. That led to a scoped-down application tracker and Verama+, a blog/forum hybrid for freelancers.

Outcome

What Was Done

Built a design system from scratch, then two full redesigns from it for different audiences. SAT: seven recruiter interviews shaped navigation testing, document handling, and a companion mobile app for COVID-era conditions. Verama: survey findings shaped a scoped application tracker and a new community feature, built alongside other designers extending the same system.

What Could Be Improved

The biggest gap, true for both: no analytics access — no way to see how people actually used either product after launch. Every decision was grounded in real research, but none of it was checked against real usage afterward.

Beyond that: ongoing usability testing on SAT instead of stopping at initial research, plus reporting features giving recruiters pipeline visibility over time. On Verama, validating the full Kanban board once engineering could support the automation.