Rebuilding the foundation of a platform so it could keep growing.
Prelude needed to evolve. New features had to be added, existing ones redesigned, but every change became harder than it should have been. The reason wasn't immediately obvious: over time, almost everything had ended up inside forms.
Study settings, permissions, integrations, reporting, notifications, even feature configuration, had gradually been pushed into the form builder. Forms were meant to collect subject data, but they had become the control center for the entire study. Configuring a study was slow. Even something as routine as adding a calculation often required custom code. The problem was that the product no longer had a clear place for new functionality.
Everything started with one architectural decision: forms would only collect subject data. Each part of the product got one clear job. Forms went back to collecting subject data and nothing more. Study settings became their own section. Feature settings became another. Reporting, exports, medical coding, imports, and group data entry moved under Data Management. Notifications became a dedicated page.
This decision affected everything that followed. The builder, navigation, reporting, and every new capability depended on getting the product architecture right.
Abstracted structure diagram. Real product screens are under NDA.
Once forms had a single responsibility again, I could redesign the builder itself.
The builder and the final subject record behaved differently, making forms difficult to understand and even harder to maintain. Adding calculations or validation rules meant jumping between different parts of the interface, and many configurations still required code. I redesigned the builder around the field itself. Every field became the place where its logic lived. Calculations, constraints, and validations moved alongside the data they controlled, making the builder predictable, easier to scan, and fully visual. Rules became searchable in a validation library.
Abstracted concept of the redesigned builder. Values are illustrative.
Study build time fell by 75%. Configuring a new study became faster because the builder no longer depended on custom code or too many clicks.
Once the product architecture was in place, new features could be designed as products of their own instead of being squeezed into existing workflows.
Pet owners previously logged into the same EDC used by study staff, with heavily restricted permissions. It worked, but it wasn't designed for them.
I designed a dedicated patient app where owners complete questionnaires, review study documents, and sign eConsent. Everything syncs directly to the EDC, reducing site visits and improving data quality.
Adverse event and concomitant medication coding no longer lived inside forms. I designed a dedicated workspace connected directly to each subject's record, giving coders access to the clinical context they needed without navigating the rest of the study.
The reporting system combined standard ready-made operational reports with configurable ones built from the same underlying components.
Instead of creating a new inventory model for every study, I reduced them to three reusable patterns that covered the workflows clients actually needed.
RTSM was the hardest problem I worked on, and the one I'm proudest of. Every study used a custom inventory model. My goal was to reduce those into a small set of reusable patterns that could support many different trial designs.
Engineering saw the problem differently than I did. We spent hours sketching different models, walking through edge cases, and pressure-testing each one against real study scenarios. By the end, neither of us was looking at the solution we started with.
We combined the strongest parts of both approaches into a model that was simpler to build, easier to understand, and flexible enough to support different inventory workflows. We walked it through clients with very different inventory processes. Each could map their study without introducing another custom inventory model.
I care more about finding the right solution than defending my own. The final inventory model combined ideas from design and engineering, then held up across very different client workflows.
The architecture became the foundation for new capabilities across the platform, from ePRO and eConsent to Group Data Entry, Medical Coding, and RTSM, with room for future products to fit naturally into the same system.
Study build time reflects the builder redesign. The broader figures are outcomes Prelude reports publicly once teams are on the rebuilt platform, drawing on the build and data-capture flows I designed.
The lesson I keep: the highest-leverage design work is often invisible. Fixing the foundation is what made every feature after it cheaper to build, and it only happened because design and engineering were shaped together.