Systematic literature review tools that handle 500+ PDFs without losing citation context

Running a systematic review and we've accumulated ~500 PDFs across 3 databases (PubMed, arXiv, IEEE). The problem isn't finding papers — it's keeping track of: - Which papers cite which (beyond what citation graphs show) - Methodological differences between studies with similar titles - Extracting effect sizes / metrics into a structured comparison table We tried Zotero + PDF metadata extraction, but it falls apart for preprints and conference papers with non-standard metadata. Paperpile was better for collaboration but worse for programmatic extraction. Anyone using tools like Semantic Scholar API, Scite, or custom RAG pipelines for literature review management? Specifically interested in: - Automatic method extraction (e.g. "this is a randomized controlled trial with N=200") - Conflict detection (paper A says X, paper B says Y on the same question) - Export to structured formats for meta-analysis What's your workflow?