Speaker
Description
We present a framework for developing Large Language Model (LLM) applications that integrate with various data sources and systems, enabling advanced AI-driven capabilities. Our approach focuses on automating time-consuming/repetitive tasks that require knowledge work through incremental deployment of LLM applications, which can process unstructured information in a common sense manner. We utilize a Retrieval Augmented Generation (RAG) framework to incorporate external knowledge and in-context learning techniques, allowing our LLM applications to learn new skills and adapt to changing contexts. Our framework is built upon open source tools, which provide a scalable and flexible platform for developing and deploying LLM applications locally in our dedicated GPU infrastructure. We demonstrate the capabilities of our framework through various Comprehensive Nuclear-Test-Ban Treaty Organization use cases, including purpose-built AI assistants, coding assistants, and research assistants such as a paper reviewer and a plagiarism detector.
| [email protected] |
