My AI system for biomedical questions with traceable evidence.
BioEvidence AI turns a clinical or biology question into a concise answer backed by PubMed citations. This public demo shows the five systems I submitted to the TREC BioGen 2025 shared task.
A shared-task demo for evidence-grounded biomedical answer generation.
TREC BioGen Task B asks systems to answer biomedical questions using retrieved evidence. This demo exposes the public submitted outputs, not a live model endpoint.
The shared-task runs compare sparse BM25 retrieval with dense MedCPT reranking under narrow and wide evidence budgets.
Three useful signals from the official results.
Built by Ganesh Chandrasekar.
I am an ML and NLP engineer researching evidence-grounded biomedical question answering, retrieval, and trustworthy LLM evaluation at Concordia University.
Thanks to CLaC Lab, the TREC BioGen organizers, and the shared-task evaluators. Additional thesis systems will be added after defense and publication.