← Back to Work
Ingest — extract text from uploaded PDFs. Analyse — a LangGraph workflow evaluates each paper against your query and scores relevance. Synthesise — a second node summarises the matches into a final verdict with reasoning. Stateful two-node graph: per-paper analysis → cross-paper summary. Low-temperature, structured prompts for consistent relevance scoring. Per-paper error isolation and large-context handling for long documents.
( Case Study · 2025 )
AI Research Agent
Overview
A retrieval-augmented (RAG) platform that turns a pile of research papers into an interactive, queryable knowledge base — ask a question, get a source-grounded answer.
How it works
Stack
Python · LangChain + LangGraph · Groq (Llama 3.3) · Sentence Transformers + FAISS · Streamlit