About This Project
Overview
ProtoML automates the process of keeping up with ML research: Groq and Gemini agents pull new papers from arXiv, score and debate which ones matter, then process the PDFs into semantic embeddings and structured, readable summaries. Results land in Discord automatically, turning a manual literature review into a standing pipeline.
Motivation
I wanted to stop missing relevant ML papers buried in the daily arXiv flood, so I built a pipeline that reads and triages them for me instead of the other way around.
Lessons Learned
Letting two different LLM agents debate which papers matter produced noticeably better curation than a single scoring pass — disagreement between models turned out to be a useful signal, not just noise to average away.