ORA - An Ovarian Research Agent

In 2024, the AthenaDAO Science and Deal Flow team started a collaboration with Arhkai (formerly known as CoopHive), founded by Levi Rybalov, to build the first field-focused agent for ovarian aging. Our team curated a set of papers, and prepared questions-answer pairs requiring highly specialized knowledge in ovarian aging. Dr. Guilherme Gatti, a neuroscientist working on computational biology at Harvard Medical School, and AthenaDAO Science & Deal Flow member, led the endeavor. 

Arkhai provided a reproducible RAG pipeline that was intended to find the best combination of document converters, chunking strategies, embedding models, reranking strategies, hybrid search strategies, cross encoders, and more. It did this by creating a reproducible knowledge graph tracking every input and output of the RAG pipeline, so that any resulting database could be reproduced easily.

After running the papers through the pipeline and querying LLMs accessing the RAG databases, the next stage was vetting and curation by a field expert, for which we brought on Dr. Joshua Johsnon, one of the leading researchers in working on ovaries today. 

We were working on agentic AI from a female health angle but we were ultimately unable to bring the project to its fullest potential. Building datasets that can feed AI from a sex-specific perspective requires more attention and resources than it currently gets. As it stands today, the biggest bottleneck for AI in this domain is not just getting access to high quality data, but also getting the talented teams and the resources needed to develop it. 

These are the kinds of experiments we must be exploring more.