Get in touch
Get in touch
The first step in clinical trial site selection—identifying which sites and investigators might be a match for a given indication—ran through a manual analyst queue that took days.
Advarra and Aditi partnered to change how that step works: enabling sponsors to access relevant site and investigator intelligence directly, in seconds, and freeing analysts to focus on the judgment work that requires them.
An important early question in site selection is: which sites are even candidates for a trial?
At Advarra, answering that question started with an analyst. Every sponsor request entered a queue. An analyst manually mapped the submitted condition across thousands of clinical indications and approximately 120 analytical models and then prepared the results through Tableau and Excel.
The bottleneck was not the analysis itself. It was that this first step had to be completed manually, one request at a time. And it wouldn't scale.
As demand grew, scaling the service meant one of two things: add more analysts—more people, more hours and more queue—or change how the work was done. Advarra chose to change it: to evolve SiteID from a tool that served analysts into one sponsors could use directly, without giving up the human judgment that matters.
The change came down to one principle: automate what repeats, keep people on what doesn't.
The first step of site selection—mapping a condition to candidate sites—is repeatable. So Advarra moved that first step into the platform with Aditi's support. Sponsors now run that screening themselves, in seconds. Analysts step in where judgment is required, not for every request.
Two workstreams ran in parallel from day one—the governed data foundation and the self-service application—so Advarra could validate the direction early and steer, without waiting for the complete data layer to be ready.
AI accelerated how Aditi engineered the platform; automation changed how the platform works. Aditi used Claude and Claude Code to accelerate specification, code generation, pipeline development, testing and documentation. Inside the platform, indication mapping runs through a separate automated capability—it does not depend on the Claude tools used by Aditi during engineering.
AI set the pace, not the standard: Aditi engineers reviewed every output, refined the business logic and stayed accountable for every production-bound decision.
Advarra's governance requirements shaped delivery from the start: Advarra One SSO, governed data pipelines and full traceability, built to stand up to Advarra's own processes.
A sponsor submitted a medical condition and waited. An analyst connected the condition to the appropriate standardized medical code, compared it with thousands of clinical indications, worked across approximately 120 analytical models, prepared and exported the results through Tableau and Excel, and returned the information to the sponsor. Every request depended on analyst availability, adding days before potential sites could be explored.
A sponsor enters a medical condition directly into the platform. The platform automatically connects that condition to the relevant information and surfaces candidate sites in under two seconds. Routine mapping no longer passes through an analyst for every request. Analysts focus their expertise on exceptions, interpretation and decisions requiring human judgment.
Within seven workable days, Advarra had a working prototype to test and validate the proposed direction.
Following validation of the prototype, the team developed the capability into a working application in approximately six weeks.
“This team has gotten a brand-new application to the finish line in about six weeks—that's incredible productivity, especially in this industry. It has been a lot of fun to watch the team build, pivot and produce as quickly as you have.”