Max Wo

I have built a scientific literature agent that is designed to help researchers inspect AI generated answers instead of purely trusting the outputs. Researchers often need to compare claims, methods, and terminology across different papers. My project structures such papers in a way that the relationships between claims, structures, methodology, and concepts across papers can be reasoned with using clearly defined boundaries. My prototype allows researchers to ask how two papers characterize the molecular properties of RNA-targeting small molecules. Users can then open the precise passage and source metadata supporting the claim. Although this project does to claim to verify scientific truth, it demonstrates the foundation for a larger semantic knowledge graph that uses many more papers for advanced reasoning. Researchers will have the ability to compare findings across large amounts of scientific data accurately, understanding any limitations or uncertainty amongst the data.