Ganesh Chandrasekar
I build and evaluate retrieval, RAG, and language-model systems where answers need to remain connected to their evidence.
Applied AI research with an engineering foundation.
I am completing a Master of Computer Science thesis at Concordia University on evidence-grounded biomedical question answering. My work spans sparse and dense retrieval, sequence labeling, open-source LLM inference, and human and model-based evaluation.
I care about systems that can be inspected: what they retrieved, what they answered, which evidence they used, and where they failed.
From a 26.8-million-document corpus to five public runs.
For TREC BioGen 2025, I developed modular sparse and dense pipelines using BM25, Pyserini, MedCPT, cross-encoder reranking, TF-IDF/MMR, and Qwen. BioEvidence AI makes the submitted outputs and their PubMed citations easier to inspect.
Thanks to my collaborators, CLaC Lab, the TREC BioGen organizers, and the shared-task evaluators. This site presents the public submission only; additional thesis work will follow after defense and publication.
Read the proceedings paper