AI & Automation
Digital Twin Retriever App
Digital Twin Retriever is a generative AI tool that retrieves and summarizes relevant clinical cases through a chatbot interface, enabling fast access to real-world evidence.
/ Overview
Practical application
Clinical decisions often rely on limited evidence, particularly in rare conditions or complex multimorbidity. This tool surfaces similar real-world cases from the literature, providing contextual support based on over 90,000 clinical reports.
Core features
- Large-scale clinical case dataset: 90,000+ de-identified case reports sourced from PubMed Central (MultiCaRe).
- Embedding-Based Retrieval
- RAG-based chatbot
- Efficient data pipeline: Dataset and embeddings stored in Parquet format and loaded dynamically to optimize performance
- Context enrichment via scraping: Optional retrieval of external sources to expand context for complex queries
PROJECT IN DETAIL