Clinical Trial Design Platform, AstraZeneca
In 2026, I designed and built an end-to-end clinical trial design platform for AstraZeneca that rethinks how researchers define, validate, and operationalize patient cohorts. The core problem was clear: cohort creation is traditionally fragmented, slow, and dependent on specialized data teams. Researchers often translate clinical intent into rigid query logic, wait on validation, and iterate through disconnected tools. I introduced a unified system where natural language, structured criteria, and AI agents work in sync, allowing clinical scientists to move from idea to validated cohort in minutes instead of days.
At the center of the experience is an AI-assisted cohort builder that translates clinical intent into standardized concepts (OMOP/SNOMED), resolves ambiguities, and guides users through validation. The interface balances flexibility and control: users can describe cohorts conversationally or refine them through a structured criteria panel including demographics, temporal windows, comorbidities, and exclusions. The system surfaces real-time feedback such as data coverage gaps, concept expansion risks, and attrition impacts, ensuring users understand tradeoffs before running expensive queries. This tight feedback loop reduces errors and builds trust in the underlying data model.
A key innovation is the agent-based architecture layered throughout the platform. Specialized agents handle cohort construction, data quality validation, feasibility estimation, and evidence generation. These agents operate transparently within the UI, surfacing decisions, flags, and recommendations directly in context rather than as black-box outputs. The activity feed and audit trail provide full visibility into system actions, enabling collaboration across clinical, data science, and regulatory teams while maintaining compliance and traceability.
Beyond cohort creation, the platform extends into a broader ecosystem including a federated dataset marketplace and an agent catalog. Researchers can discover datasets, evaluate availability and compliance, and apply them directly within workflows. The result is a vertically integrated system that compresses the entire trial design lifecycle into a single interface. By combining AI-assisted intent parsing, structured clinical modeling, and agent-driven automation, the platform significantly accelerates trial design while improving accuracy, transparency, and scalability.