AI Development and Testing Service

The agency is seeking a vendor to provide artificial intelligence development and testing services. The primary goal is to establish AI-enabled scientific discovery frameworks that use multimodal clinical data to generate hypotheses around patient health and care, rather than just prediction models. The agency also requires world-class dataset generation frameworks to encourage the creation of curated, interoperable, AI-ready datasets that can serve as a shared resource across clinical AI projects. The scope of work covers several distinct frameworks to support clinical AI adoption and data integrity. The vendor must establish training frameworks for developing modules or curricula that support clinician AI literacy and promote adoption into clinical practice. To ensure data quality, the agency requires frameworks to minimize errors during model training, address data leakage where training data contains the target prediction information, and manage the data lifecycle to reduce drift and prevent the reintroduction of errors. Additional requirements include frameworks to quantify uncertainty stemming from data or models, harmonization frameworks to ensure multimodal data and models can be utilized together, and transparency or explainability frameworks to increase user trust in AI decision-making. Eligible bidders must be onshore organizations based in the United States. The performance of all requested work will take place offsite, and the agency is currently looking for proposals to determine the budget.