PM Accelerator — MindGym

Designing an AI-powered platform that helps jobseekers navigate applications and interviews with confidence

MindGym is an AI-powered platform built to support jobseekers through the job application and interview process. I joined as AI UI/UX Designer at PM Accelerator, working within a cross-functional team of 13 — PMs, designers and engineers — to take the product from research through to high-fidelity UI.

Conducted and synthesised research across 50+ user interviews alongside existing databases to identify jobseeker needs, behaviours and pain points. Three key themes emerged consistently: anxiety around applications, lack of personalised feedback, and difficulty preparing for interviews under pressure.

The finding that reframed the product: job searches stall through behavioural failure, not skill gaps — and behaviour only changes once mindset does. That moved the brief from "help users do more" to "find where this user is actually losing, and change how they feel before they act." Anxiety became the daily mood check-in; lack of feedback became the confidence delta; interview pressure became the timed session.

The core challenge was making an AI-driven product feel human and supportive rather than cold and transactional. AI platforms often overwhelm users with capability — MindGym needed to feel like a calm, intelligent coach rather than a powerful tool demanding to be configured.

Every screen had to balance three things simultaneously: clarity (the user always knows where they are and what to do next), warmth (the tone and visual language never feel clinical), and adaptability (AI-driven content changes per user, so the design system had to hold together across infinite variations of content).

The product's own principle set the bar: a generic meditation app says "breathe and relax" — MindGym names the company, the room, the panel, the first question. The specificity is the product, which meant no screen could fall back on placeholder warmth. Every state had to hold a real name, a real number, a real next move.

Three decisions carried most of that weight: a serif display face for headings so the product speaks rather than reports, a deep teal used only where the user should act next, and a single warm amber reserved for anything time-sensitive — so urgency reads as a colour, not an alarm.

Contributed to and maintained a component-based design system throughout — building variants, auto-layout structures and shared libraries to ensure consistency as the product scaled across multiple screen types and user journeys.

The system had to account for AI-generated content of variable length, meaning every component needed to be genuinely flexible, not just theoretically so. This discipline — designing components that hold together under real-world content variation — was one of the most valuable parts of the project.

By handoff the library covered sixteen button states, five input states, the three-state card, the six-stage hiring funnel, the scored metric tile, the dropdown and FAQ patterns, and the nav in active and default — all on DM Serif Display over DM Sans.

Worked within an agile cross-functional team of 13, participating in design reviews, presenting work to stakeholders, incorporating structured feedback and iterating rapidly. The pace of the project required clear communication about design decisions — why something was designed the way it was, not just what it looked like.

Used Figma AI throughout to prototype faster and explore multiple design directions simultaneously without sacrificing quality or introducing inconsistency.