Nodoz
AI research intelligence platform that recommends ML architectures for scientific datasets using cross-domain similarity over 30,000+ papers.
View project →The central observation behind Nodoz is that the most productive discoveries in science tend to happen not within fields but in the transfer of methodology between them. MRI borrowed its core technique from astrophysics. Information theory migrated into genetics. These transfers are sparse and largely accidental at human scale. They require someone to be standing at the intersection of two fields at the right time and notice a structural resemblance.
Nodoz is an attempt to make that process systematic. The platform stores the core structure of data-model pairs across 30,000+ scientific papers and answers a specific question for researchers: given your data, what methodology should you try, including from fields you have never worked in?
The underlying framework is rooted in the philosopher Mary Hesse’s work on analogical reasoning across scientific disciplines. Hesse Research, the parent entity I am building, takes this as its founding principle. Nodoz is its first product.