EdgeRAG

Interactive Demo Available

• Actively developing an edge-optimized RAG system for financial analysis, focusing on retrieval accuracy, latency, and response grounding.
• Iterating on document ingestion and embedding strategies, including chunking heuristics and metadata-aware retrieval for financial text.
• Exploring evaluation methodologies for RAG systems by comparing standalone LLM outputs against retrieval-augmented responses for factual consistency.
• Experimenting with deployment constraints at the edge, including memory usage, model size trade-offs, and inference performance.
• Investigating future extensions such as real-time data integration, hybrid dense–sparse retrieval, and robust prompt orchestration.