An AI-powered ATS that thinks like a recruiter.
A ground-up product design for the flagship applicant tracking system at the company I work with — where AI does the heavy lifting of screening and ranking, and recruiters stay firmly in control. The largest product the org has shipped to date.
- Role
- Lead Product Designer — sole designer
- Timeline
- 2024 — Present
- Team
- 2 PMs · 6 engineers
- Platform
- Web app · desktop-first
- Scope
- Research → UX → UI → Design system
- Tools
- Figma · FigJam
Overview
This is the biggest product the company has built — an internal applicant tracking system (ATS) used by recruiters to run hiring end-to-end, from a fresh job opening to a signed offer. I joined as the sole designer and owned the experience from research through a production-ready design system.
Traditional ATS tools are glorified spreadsheets: they store applicants but leave the hardest work — actually reading, comparing, and judging hundreds of candidates — to overloaded recruiters. The brief was to change that. Put AI at the center of screening and ranking, but design it so recruiters trust it, understand it, and stay in charge of every decision.
The problem — and how we solved it.
Recruiters were drowning. A single open role could pull in hundreds of applicants, and the only way through was manual — opening CV after CV, eyeballing keywords, and making fast, inconsistent calls under pressure.
The old tooling made it worse, not better: it tracked candidates but offered no help judging them. So every pain got a deliberate, AI-assisted answer — paired below.
Hours lost to manually screening resumes for every role.
AI parses every CV into a structured, comparable profile in seconds.
Keyword filters quietly rejected strong, non-obvious candidates.
Fit scoring weighs the whole profile against the role — not just keywords.
No consistent, explainable way to compare applicants side by side.
Every score ships with plain-language reasoning a recruiter can defend.
A fragmented workflow — sourcing, screening, scheduling, notes — scattered across screens.
One pipeline unifies sourcing, screening, scheduling, and notes in a single view.
Slow time-to-hire let strong candidates go cold.
Ranked shortlists surface the strongest people first, from day one.
Design goals
Cut screening time dramatically without lowering the bar on quality.
Surface the best-fit candidates first — with reasons a recruiter can defend.
Keep humans in control: AI recommends, people decide.
Unify the whole hiring funnel into one coherent, learnable workflow.
Ship a scalable design system the team could build on for years.
Process — brief to shipped.
Three phases, one designer, end to end — no handoffs, no lost context. The work leaned hard on talking to real recruiters and shadowing live screening before a single screen got drawn.
The full design journey — drag the canvas to explore it like a map.
- Discover · 01
Discovery & recruiter interviews
Sat with recruiters and hiring managers, shadowed live screening sessions, and mapped where time actually went. The pain wasn't tracking candidates — it was judging them.
- Discover · 02
Mapping the hiring funnel
Reconstructed the end-to-end journey — job intake, sourcing, screening, interviews, offer — to find the seams where work fell through the cracks and AI could earn its keep.
- Design · 03
Information architecture
Restructured a sprawling tool into a clear spine: jobs, candidates, pipeline, and analytics — each a place with one job, navigable without a manual.
- Design · 04
Flows & wireframes
Designed the core loops in low fidelity first, pressure-testing the screening and ranking flow with recruiters before committing to any visual direction.
- Design · 05
Designing the AI patterns
The hardest, most interesting part: how to present a score, an explanation, and a recommendation so a recruiter trusts it. Built reusable patterns for confidence, explainability, and override.
- Design · 06
Visual design & system
Turned the flows into a polished, accessible UI and codified every decision into a design system — tokens, components, and AI patterns the team could reuse.
- Deliver · 07
Validation & handoff
Tested prototypes with the people who'd live in the tool daily, iterated on the rough edges, and partnered closely with engineering through build.
Inside the product
Eight areas carry most of the weight. Each was designed to take work off the recruiter without taking away their judgment.
AI resume parsing
Turns messy PDFs and CVs into structured, comparable profiles in seconds — skills, experience, and education, normalized.
Match scoring & ranking
Every applicant gets a transparent fit score against the role, so the strongest candidates rise to the top of the pile automatically.
Why this candidate
Plain-language reasoning behind every score — skills matched, gaps flagged — so recruiters can trust it and defend it.
Candidate pipeline
A drag-and-drop Kanban across every hiring stage, so a recruiter always knows what's moving and what's stuck.
Smart candidate profile
One view for everything: resume, scores, AI summary, notes, and activity — no more hopping between tabs.
Job intake & posting
Guided job creation that captures real requirements and feeds the matching model the signal it needs.
Interview scheduling
Availability, panels, and reminders handled inside the platform — no calendar ping-pong.
Recruiter analytics
Funnel health, time-to-hire, and source quality at a glance, so the team can see what's working.
Key screens
A look at the core surfaces. (Placeholders for now — real screens drop straight in.)
Design system
Because this is the org's biggest, longest-lived product, the system mattered as much as any screen. The palette is deliberately quiet — a near-neutral, paper-like canvas so candidate data and AI signals carry the color. One decisive blue drives every primary action, and a tight set of semantic colors tells the story of a candidate's journey through the pipeline.
AaBbCcDdEeFfGgHhIiJjKkLlMmNnOoPpQqRrSsTtUuVvWwXxYyZz 0123456789 & @ # % ( ) ! ?
Surface & borders
Text & icons
Accent & links
Gradients
Candidate status
Explainable by default
No black boxes. Every AI score is paired with the reasoning behind it.
Human-in-the-loop
AI recommends, recruiters decide. Every suggestion is easy to override.
Confidence, not certainty
The UI shows match strength honestly and never hides uncertainty.
Progressive disclosure
A dense, powerful tool that stays calm — depth revealed only on demand.
Outcome & impact.
The redesign reframed the ATS from a passive system of record into an active hiring partner. Recruiters spend their time judging the right shortlist instead of digging through the whole pile — and they can explain every call they make.
It's now the company's flagship product and the backbone of how the team hires. The design system that came out of it continues to speed up everything built around it.