Clinical operations intelligence belongs on the Lakehouse
The Site Feasibility Workbench, an open-source Databricks App, now enables clinical trial site selection entirely within the Databricks workspace, eliminating external API calls and synchronization pipelines. This solution addresses the architectural challenge of disconnected clinical operations data, improving enrollment target attainment with TA-segmented LightGBM models and auditable SHAP-driven explanations.
* What it is: The Site Feasibility Workbench is an open-source Databricks App that runs clinical trial site selection entirely within the Databricks workspace — combining ML-driven site scoring, Lakebase for operational state, and AI/BI Genie for natural language data access, with no external API calls or synchronization pipelines. * The challenge it solves: 37% of investigator sites miss enrollment targets, and the root cause is architectural — clinical operations data and the applications that use it live in disconnected systems, forcing decisions into spreadsheets and creating integration overhead, credential sprawl, and synchronization lag that erodes trust in the data. * Results and outcomes: TA-segmented LightGBM models trained on your own CTMS, EDC, and IRT history — not industry averages — produce scores that improve as your portfolio grows, with SHAP-driven explanations stored as governed, versioned Delta tables. Every prediction carries SHAP-driven attribution stored as a governed Delta table, making model rationale as auditable and versioned as the score itself.