Summarization RAG System

Context: Global commodity pricing and intelligence platform serving 600+ editorial staff.

Challenge: Editorial team needed to extract key insights from 145,000+ historical articles and 800 daily market-moving stories to identify trends and generate actionable intelligence – all without manual reading.

Solution: Built an AI-driven text generation system using:

• Gemini LLM + LangChain for retrieval-augmented generation
• AWS S3 and OpenSearch for storing and indexing historical articles
• Integrated R Shiny frontend via APIs for editorial workflow
• Deployed on Posit Connect for seamless access

Outcome: Enabled the editorial team to rapidly surface market trends, generate summaries, and produce data-driven intelligence – dramatically reducing manual research time.