SANYAMWADHWA.TECH :: PROJECT :: PAPERMIND --- SUMMARY ---TITLE............. PaperMind — Faithful Research-Paper IntelligenceDATE.............. June 2026CATEGORY.......... ML & ResearchPROBLEM........... Abstract-only summarizers miss what a paper actually shows. PaperMind processes the full PDF, every section, table, and result, then grades its own output for faithfulness before returning it.WHAT IT DOES...... LangGraph-orchestrated pipeline that turns full academic PDFs into structured intelligence: map-reduce summaries, typed entity and results extraction, LLM-judge quality grading, and a knowledge graph of how papers relate. --- RESULT ---METRIC 1.......... ROUGE-1 F1: 0.64 vs abstractsMETRIC 2.......... 90/90 tests passingMETRIC 3.......... $0 API cost (free tiers) --- WHAT I BUILT ---HIGHLIGHT 1....... LangGraph map-reduce pipeline covers the whole paper: every section is summarized concurrently, then synthesized into a 300 to 450 word summary with findings, contributions, and limitations.HIGHLIGHT 2....... An LLM-as-judge grading node scores faithfulness and specificity and retries weak results automatically, with cascading failover across Gemini 2.0 Flash, Groq Llama 3.3, and local Ollama Qwen2.5. --- HOW IT WORKS ---STEP 1............ React 18 + Vite FrontendSTEP 2............ FastAPI (async)STEP 3............ LangGraph EngineSTEP 4............ Supabase (pgvector)STEP 5............ Gemini 2.0 FlashSTEP 6............ Groq Llama 3.3STEP 7............ Ollama Qwen2.5 (cascading failover) --- FEATURES ---FEATURE 1......... Map-reduce summarization: prepare, map sections concurrently, synthesize, then gradeFEATURE 2......... Typed entity and quantitative-results extraction via Pydantic schemas, from prose and real PDF tables; no regexFEATURE 3......... LLM-as-judge quality control: weak summaries loop back and get retried automaticallyFEATURE 4......... RelationAgent labels paper-to-paper links (extends, replicates, contradicts, shares-method) into a pgvector knowledge graphFEATURE 5......... Domain-agnostic across biomedicine, physics, and ML papersFEATURE 6......... Benchmark harness runs the full pipeline against real arXiv papers: 100% success rate, up to 2.5x parallel speedup, and it caught a real figure-extraction bugFEATURE 7......... Runs entirely on free LLM tiers (Gemini, Groq, local Ollama) with rate limits absorbed by concurrency caps and backoff retries --- STACK ---STACK............. LangGraph, FastAPI, React, Supabase pgvector, Gemini 2.0 Flash, MinerU --- LINKS ---CODE.............. https://github.com/SanyamWadhwa07/PaperMindHUMAN PAGE........ https://sanyamwadhwa.tech/projects/papermind