September 15, 2026
The National Library of Medicine (NLM) is launching the Semantic Precision for AI Retrieval of Knowledge (SPARK) Challenges. These prize competitions, totaling $500,000 across two areas, will advance AI-enabled biomedical literature search and data discovery in support of NLM’s core mission. SPARK invites AI engineers, data scientists, researchers, startups, and nonprofits to transform how the world finds, synthesizes, and builds on biomedical knowledge.
PubMed® and PubMed® Central (PMC) now index more than 35 million biomedical abstracts and full-text articles. The NIH Database of Genotypes and Phenotypes (dbGaP) holds thousands of clinical study variables recorded in inconsistent, non-standardized shorthand. Navigating these resources efficiently and reproducibly remains one of the defining challenges of modern biomedical research.
SPARK challenges innovators to build AI systems capable of context-aware, meaning-driven discovery: systems that don’t just match keywords but understand scientific evidence, prioritize reproducibility of research findings, and accelerate cross-disciplinary discovery.
PubMed/PMC® Challenge Tasks
- Known-Item and Provenance Retrieval: Retrieve the specific publication answering a defined biomedical question.
- Open-Ended Exploratory Search: Deliver a synthesized, citation-backed natural language response to broad research questions.
dbGaP Challenge Tasks
- Cross-Ontology Precision Mapping: Map non-standardized dbGaP study variables to LOINC®, RxNorm®, SNOMED-CT®, and the Unified Medical Language System® (UMLS®).
- Natural Language Discovery Interface: Let researchers surface relevant clinical studies from dbGaP using plain language queries.
SPARK is open to individuals, teams, companies, academic institutions, and nonprofits. Eligibility details and submission requirements are available at NIH Challenges and Prize Competitions.
Registration opens Sept. 15, 2026. Deadline for solution submissions will be Jan. 15, 2027, and result submissions are due by Jan. 30, 2027.
