IAEA (International Atomic Energy Agency) · Sep 2022 - Apr 2023 · 8 months
Designing the analytics platform behind global nuclear safeguards verification
As lead UX/UI designer, I designed the IAEA's sample evaluation platform end-to-end - dashboards and data tracking, a query builder and plotting toolkit for deep analysis, and a portable field app for collecting samples - replacing a decades-old, spreadsheet-style interface, almost entirely without direct access to real users or real data.
Team
Me, our project manager (the only one cleared for IAEA HQ access), and developers
Scope
Dashboards & data tracking, a data constructor with query builder and plotting, and N Portal, a portable field app
Constraint
No direct access to SMEs or real data - requirements came through our PM
Duration
8 months, Sep 2022 - Apr 2023
The problem
Before this project, the platform looked like software from the 1990s - dense grids of cells and numbers, no real visualization, and none of the capability scientists actually needed: no query builder, no plotting tools, no way to bring samples collected in the field into the system directly. This platform is used to detect and locate microscopic nuclear particles in environmental swipe samples, verifying state compliance with non-proliferation treaties - work with real geopolitical weight if a result gets missed or misread. I designed the replacement from the ground up. The added constraint: this was sensitive enough that I never had direct access to the scientists who'd use it, or to real sample data.
Process
Modernized dashboards and data tracking before touching the specialized tools
The existing system looked like it hadn't changed since the 90s - dense cells and numbers, no visualization. I started with the dashboards and data tracking screens scientists used most often, since that's where the daily value was, before moving on to the more specialized, lower-frequency tools.
Built the data constructor, query builder, and plotting tools from nothing
None of this existed before: a way to construct custom data views, query the underlying sample data directly, or plot results for analysis - including SEM (Scanning Electron Microscopy) data used for particle screening, morphological and elemental analysis, and selecting targets for secondary analysis. I designed these as one connected set - build a query, see it as data, see it as a plot - instead of three disconnected tools bolted together.
Designed N Portal as a portable, field-ready app, not a shrunk desktop screen
N Portal needed to work for someone adding samples on-site, not at a desk. I designed it as its own portable experience built around the actual field-collection workflow, instead of just squeezing the desktop data-entry screens onto a smaller device.
Selected screens
Key decisions
Designed for cross-functional context in a dense, specialized interface
Analysts needed to see how a result connected to related samples, prior findings, and reference data - not just the one record in front of them. I designed the review and analysis screens to surface that context directly, instead of leaving analysts to hold it all in their head across separate screens.
Automated the handoffs instead of just digitizing them
Much of the old process was slow not because individual steps were hard, but because handing data from one stage to the next was manual and insecure. I designed the workflow to move data between stages automatically and securely, instead of just putting a nicer interface on the same manual handoffs.
Outcome
489 screens
designed and deployed across dashboards, the query builder & plotting toolkit, and N Portal
100% audit traceability
across analytical outputs - a hard requirement for software used in nuclear safeguards verification
Automated handoffs
data moved between stages securely and automatically instead of through manual handoff
Figures are drawn from a public case study Rokolabs (the agency I worked through on this engagement) has since published.
What I’d do differently
“I never sat with the scientists who'd actually use this, and for a while that felt like a gap I should be compensating for. It wasn't. Our project manager's daily, firsthand knowledge of how IAEA HQ actually worked was more reliable than a handful of remote user interviews would have been anyway. The real lesson was trusting that proxy relationship fully, instead of treating it as a lesser substitute for research I was never going to get.”