Two and a half years embedded at the institution that grades millions of diamonds annually, across nine global laboratories in thirteen countries. My work was the toolset that made that throughput possible.
Under NDA Specific metrics aren't public. Structural proxy: six product surfaces shipped across grading, mobile, and trade-facing contexts; the toolset remains in production daily use across nine global laboratories.
Numbers Available on a callAll of this is publicly sourced — none of it is proprietary. It is the institutional gravity that makes designing for GIA a different problem than designing for a consumer app, and it is the lens I want a hiring manager to apply to everything below.
Scale. GIA grades millions of diamonds every year and issues millions of reports to match, across nine global laboratories in thirteen countries. Every one of those reports is the output of an internal workflow I contributed design to — intake, photography, analysis, grading, cross-verification, inscription, publication. At that scale, a small improvement to the grading step compounds across tens of thousands of stones a month.
Standard-setting. GIA created the 4Cs and the International Diamond Grading System in 1953. Those standards are now the universal language of the diamond trade — used by retailers, auction houses, insurers, and appraisers in over 100 countries. The institution has been setting the standard since 1931. Designing its tools means designing inside a vocabulary the whole industry already shares and will not let you redefine.
Consequence. A GIA report is the difference between “a diamond” and “a documented asset.” The grades those internal tools help produce go on to appear in insurance claims, estate documents, and auction records. The user on the other side of the interface is a credentialed gemologist staking professional reputation on getting it right — thirteen countries’ worth of them, working the same systems to the same standard.
Designing for professional reputation is a different problem than designing for delight — and the difference shows up in what counts as “good.”
The institution’s scale is the reason the design work is non-negotiable. Millions of diamonds are graded each year by teams of gemologists working through internal systems — intake, analysis, grading, cross-verification, publication — for stones that will go on to carry real financial and legal weight. There is no margin in those systems for an interface that requires interpretation.
That reframes what “good design” means here. In a consumer product, a moment of friction costs a little engagement. In a grading workflow, a moment of ambiguity is a potential error in a document an insurer will rely on. The job is not to make the tool feel pleasant. The job is to make the tool feel certain — so that a grader working at speed, for hours, never has to stop and ask the interface what it means.
Designing toward certainty rather than delight is the through line of the entire engagement. It changes what gets prototyped, what gets tested, and what counts as a finished component. It is also the single most useful thing I can tell a hiring manager about how I work: I optimise for the standard the domain actually holds, not the standard a design portfolio rewards.
Six product surfaces, and the engagement’s real difficulty is that they are not one problem at six scales — they are six genuinely different problems that happen to share an institution. On some I held end-to-end ownership, research through to shipped UI. On others I was the UX advocate between GIA stakeholders and third-party teams — the person whose job was to keep the user present in decisions that would otherwise have been settled without them. Knowing which mode a surface needs, and not defaulting to the one that flatters a portfolio, is most of the work.
Systems used by GIA staff across global campuses and grading laboratories — full creative ownership. The precision requirement was highest here: graders and researchers have no tolerance for an ambiguous interface, and the primary grading UI is judged by whether it disappears at the speed they work.
Design contributions to GIA’s mobile application, which serves two journeys at once: someone buying their first piece of jewellery, and a trade professional verifying a report in the field. One app, two users with opposite expectations — the design challenge is serving both without compromising either.
Client-facing tools for the trade, surfaces for GIA subsidiaries, and print — the grading report itself, where typography and hierarchy carry legal weight. Across all of them my role ranged from full ownership to UX advocacy, holding the user’s case inside multi-party decisions.
My primary role at GIA was UX/UI design: research, information architecture, interaction design, high-fidelity UI, and QA through to production. For a significant period I carried that alongside the responsibilities of interim Project Manager — owning sprint ceremonies, the Jira board, backlog grooming, and coordination across design, engineering, and QA, until a permanent PM was hired.
Taking on the second role wasn’t a promotion and it wasn’t ancillary to the design work. It was the team recognising that cross-functional coordination was on the critical path — and that I already understood how design decisions moved through the full product-development cycle. The trust it reflected was earned before it was formalised.
What changed for me as a designer: I stopped thinking of the handoff as the end of my work and started treating it as a design surface in its own right. Running the board taught me exactly what an under-specified ticket becomes by the time it reaches production — and that lesson now sits upstream, in how I document and annotate, rather than downstream in implementation review.
In a domain where error has consequences, an ambiguous state is not a creative choice — it is a defect.
Selected design calls from the internal enterprise-platform work. Each one had a credible alternative; each is recorded with the alternative it beat and the reason.
The primary grading UI uses dense layouts, keyboard-first flows, and minimal hand-holding. The rejected alternative — progressive disclosure that accommodates both newcomers and experts in one surface — would have compromised both. The primary users are credentialed gemologists grading for hours daily; the institution already runs training. Onboarding belongs in the training surfaces, so the grading tool could serve the expert case without apology.
WCAG 2.1 AA was treated as a design requirement from the first artifact — contrast, semantic hierarchy, and focus management built into every component as it was drawn. The rejected alternative, a compliance pass at the end, reliably produces a retrofit that satisfies an audit and still fails a real user. Built in from the start, it costs nothing extra and changes the result.
In a domain where error has consequences, an ambiguous state is not a creative choice — it is a bug. Every label, every state, every interaction is designed to confirm that the system understands exactly what the grader is doing and requires no interpretation from them. “Is this visually interesting” gave way to “does this remove any possible ambiguity.” When the two answers diverge, the second one wins.
Running sprints while designing made the cost of a vague handoff impossible to ignore: vague annotations become engineering judgement calls, missing edge cases become bugs, a component without a loading state gets one invented by someone who didn’t design it. So annotation layers and edge-case documentation are now part of the design process — not a cleanup step. The handoff should leave no decision the designer would have made differently than the engineer.
Every interactive component, eight states drawn: default · hover · focus · active · disabled · loading · error · empty.
Precision over delight. I came into this role with the intuition that good design should feel effortless. That is still true — but I had to recalibrate what “effortless” means for a user whose attention is professionally deployed. For a GIA grader, effortless means frictionless precision: every label, every state confirming the system understands exactly what they are doing. The standard isn’t “is this pleasant.” It is “does this require nothing of the user’s interpretation.”
Specificity in the handoff. The quality of a handoff determines how much of the design actually reaches production intact. That is not a process nicety — in a domain where the shipped tool carries consequence, an invented loading state or a misread annotation is a defect in the final product. Specificity upstream is the cheapest quality control available.
Shared vocabulary with engineering. Domain-specific terminology means different things to different teams; a misaligned definition inside a ticket creates ambiguity that surfaces in production. If I were starting the engagement again, I would invest earlier in a shared glossary, built into sprint planning — it would have saved real time I instead spent in implementation review.
The most underrated skill in an embedded role is institutional patience. Early in this engagement I had more ideas than the organisation was ready to implement. I learned to distinguish “this should change” from “this should change now” — and to invest in the second without losing sight of the first.
The designs I am most proud of from this role are not the most visually interesting ones. They are the ones where I understood a user’s workflow well enough to make a structural decision that simplified it — and where the simplification was clear enough that every stakeholder could see it immediately, without me having to explain it. Good design in a high-precision context is invisible. The win is that users think the tool is smart, not that the designer was clever.
Two and a half years in, that is the discipline I would hand to anyone starting a long engagement inside an institution: read the workflow before you redraw the screen, earn the structural changes, and let the best work go unnoticed. The full story — the surfaces, the specific decisions, the things an NDA keeps off this page — lives in the room. I am glad to walk through it there.
Two and a half years designing for graders who stake a professional reputation on every report taught me to treat ambiguity as a defect, not a style choice — a standard I now hold in every domain, whether or not the stakes are visible.
