Keeping up with the volume of new research across S2S prediction, compound extremes, and AI weather modelling is a constant challenge. This project automates that process with a lightweight pipeline that pulls new papers from journal RSS feeds, scores them against a configurable research profile using an LLM, and logs relevant hits for later review.
The screening prompt is shaped by a profile derived from my Zotero library using the zotero-mcp server, which extracts key journals, topics, and methods to ground the relevance scoring in my actual reading history.
View on GitHub →