Launching on May 15 in collaboration with cStructure and BeeKeeperAI, this challenge introduces a groundbreaking methodological approach that combines foundation Large Language Models with collaborative causal graphs within a privacy enhancing platform with a Trusted Execution Environment with confidential computing. Using cStructure’s innovative platform, participants will develop Structural Causal Models to estimate treatment effects of glucocorticoids on COVID-19 patient survival. Model validation will occur in EscrowAI’s multi-party collaboration platform, where models and data will be protected at rest, in transit, and during the computing cycle. This unique privacy preserving approach enables robust causal inference on sensitive clinical data without compromising privacy or security.
Participants will leverage LLM-assisted interfaces to identify confounding variables and determine optimal adjustment strategies, with their models evaluated against randomized controlled trial benchmarks. The challenge demonstrates how modern AI technologies can help scientists identify and estimate valuable causal insights from observational data while maintaining strict privacy protections—potentially transforming how we approach evidence generation in healthcare and beyond. Join us in this collaborative effort to advance causal inference methodology that bridges the gap between data science and medical knowledge.
- Pre-Registration opens April 15, 2025
- Challenge opens May 15, 2025
