Professional Me

Design principles

The product needed to support many users, data sources, and decisions without becoming another dense enterprise tool. I used four principles to keep the system clear while the platform evolved quickly.

• Explain the recommendation: every score needed evidence, criteria, restrictions, and visible reasoning.
• Keep people in control: AI could process volume, but the hiring team had to define what success meant for each role.
• Reveal complexity gradually: the interface should answer the immediate question first, then expose deeper detail when needed.
• Design the system, not just the screen: jobs, candidates, teams, sources, profiles, and assessments had to behave as connected objects.

Product foundation

One of the first challenges was defining how the core objects related to each other. An organization contains teams and users. Teams open jobs. Jobs receive candidates from several sources. Candidates are assessed against role-specific criteria. Different users need different permissions at each stage.

Organization canvas

The organization canvas helped administrators turn imported people and company data into a usable structure. Instead of forcing users through disconnected forms, the canvas made teams, reporting relationships, gaps, and ownership visible in one place.

This feature had to feel flexible without becoming a decorative org chart. I defined interaction patterns for importing people, placing them into teams, selecting and editing nodes, assigning relationships, and keeping the structure reliable enough to support permissions and future workforce intelligence.

Roles and permissions

Permissions were designed as part of the product experience, not just as a settings screen. Admins, recruiters, recruiter leads, hiring managers, sourcers, and interviewers needed different navigation, available actions, visible data, empty states, and collaboration options.

This mattered especially for bias-aware hiring. The interface needed to show enough context for a fair evaluation while limiting unnecessary exposure to sensitive or irrelevant profile information at earlier stages.

From job description to assessment criteria

Professional Me needed to turn a job description into criteria that could be reviewed, adjusted, and applied consistently. The system could propose a structured starting point, but the hiring team remained responsible for defining what mattered.

• Criteria sliders made relative importance visible and editable.
• Required criteria acted as explicit constraints instead of being hidden inside one opaque score.
• Basic and strict evaluation controls helped users decide how narrowly a requirement should be interpreted.
• The three most restrictive criteria were surfaced so recruiters could see which requirements were limiting the pool.

Making restrictive criteria visible

When a role produces too few suitable candidates, the problem may be the criteria rather than the talent pool. The product surfaced the three most restrictive criteria so recruiters could understand which requirements were removing the greatest number of candidates.

That turned hidden scoring behavior into an actionable design decision. Instead of lowering standards blindly, teams could adjust a specific requirement, change its strictness, or confirm that the smaller pool was intentional.

Candidate processing at scale

Bulk candidate processing involved extraction, normalization, import, and assessment. I translated that technical pipeline into a user-facing experience that communicated progress, expectations, partial completion, and recovery.

• Extract professional information from uploaded files.
• Normalize inconsistent data across resumes and documents.
• Import candidates into the correct role and source context.
• Assess every profile against the selected job criteria.
• Explain file-level failures, including cases such as documents over the supported page limit.

For one Growth Marketer role, 1,581 CVs were uploaded and 463 candidates were assessed in the recorded product snapshot. At that scale, the interface could not assume recruiters would open every profile.

High-volume candidate review

The candidate list needed to support fast scanning, ranking, filtering, status awareness, and movement between summary and detail. Scores helped prioritize attention, but the score was treated as an entry point rather than the conclusion.

The profile detail view gave recruiters the deeper evidence needed for an actual decision: strengths, gaps, criteria results, source context, and next actions. This prevented the AI output from becoming a black-box ranking.

Bias-aware evaluation

Bias-free hiring affected what information was shown, when it was shown, and how candidates were compared. The product supported both limited pre-screen views and richer detailed views so teams could begin with role-relevant evidence before reviewing broader profile context.

Source intelligence

Candidates could enter through company search, referrals, database search, the Professional Me job board, external job boards, marketplace, nationalization initiatives, and manual upload. Each source needed its own setup, activity, performance, and contribution to the pipeline.

The Sources dashboard connected channel activity to assessment quality, helping teams compare volume with actual fit and decide where to focus next.

• Where are candidates coming from?
• Which sources produce qualified people, not just more applications?
• Where is volume high but quality low?
• Which channel needs attention before the next role opens?

Referrals and external participation

Referral flows involved people who might not use the product every day. I designed the flow so contributors could understand the role, refer someone, and provide useful structured information without learning the full platform.

Playground

The Playground reduced the distance between an early hiring idea and a useful result. Users could refine a job description and explore five bias-free candidate matches before committing to a complete hiring workflow.

The product included 100 free credits, so credit awareness had to be visible without distracting from the core loop: describe the role, improve it, review matches, and continue when the value was clear.

Top Prospects

Top Prospects extended the product beyond active applicants. It helped teams discover passive candidates already present in their wider talent data and identify people worth engaging before they applied.

This required a different interaction model from a standard applicant list. The user was exploring potential, checking relevance, and deciding whether to initiate contact, so the design emphasized why a person surfaced and what the recruiter could do next.

Trust and explainability

Research participants responded positively to the clean interface and conversational AI, but they also questioned data accuracy, recommendation relevance, and how seniority or experience estimates were reached.

• Criteria-level score breakdowns made recommendations inspectable.
• Evidence was connected to claims instead of hidden behind a final score.
• Known information, inferred information, restrictions, and gaps were visually separated.
• Editable criteria kept the hiring team in control instead of treating AI judgment as final.

AI chat over professional data

The conversational layer helped users ask specific questions across information that would otherwise require opening many documents and screens. Suggested questions made the feature easier to understand, while copyable answers made insights useful outside the product.

The interaction still needed boundaries. Users had to understand whether the system was reporting a fact, making an interpretation, or suggesting a next step, and what data the answer came from.

Design system and handoff

I worked on reusable patterns for page structure, tables, filters, search, sorting, bulk actions, profile summaries, criteria controls, modals, processing states, empty states, permissions, and score presentation.

My handoffs included responsive behavior, component states, edge cases, interaction notes, and realistic data. I stayed involved during development to review builds, resolve tradeoffs, and protect consistency across the product.

What changed

The result was a more connected product experience across the recruitment journey. Large candidate pools became easier to prioritize, assessment logic became visible and adjustable, and sources could be compared by relevance rather than application volume alone.

Feature outcomes

• Organization setup became a structured workspace for people, teams, reporting relationships, roles, and permissions.
• Job requirements became a shared evaluation model through weighted criteria, required criteria, strictness controls, and restrictive-criteria feedback.
• Candidate review became more scalable through upload states, processing feedback, ranking, filtering, assessment detail, and recovery paths.
• Sourcing became measurable through source setup, source quality comparison, referrals, external participation, and passive prospect discovery.
• AI recommendations became easier to trust through visible evidence, score breakdowns, known-versus-inferred data, gaps, restrictions, and editable criteria.

What I learned

AI products need inspectability, not just simplicity. Removing every detail can make an experience feel easier, but it can also remove the information users need to trust it.

Professional Me was a systems design challenge as much as an interface design project. The work required me to connect organizational structure, candidate data, sourcing, assessment logic, collaboration, permissions, and AI into one understandable experience.