3294 - 人工智能驱动的医学研究：德克萨斯大学西南医学中心与IBM的合作.pdf

1、Orlando,FLOctober 69IBM TechXchange 2025Chinthak MuraliComputational ScientistUT Southwestern Medical Center.AI Driven Medical Research:UT Southwestern-IBM CollaborationSession ID:3294Outline Terms and definitions Vector Store RAG Framework IBM Fusion Platform Lang-graph workflow AI Avatars Live Dem

2、onstration Astrocyte Q&AIBM TechXchange|2025 IBM Corporation3IBM TechXchange|2025 IBM CorporationImportant terms and definitionsLLM An AI system trained on vast amounts of text data to understand,generate,and respond to human language in a conversational or task-oriented way.Ollama Free open-source

3、platform that makes pre-trained LLM models available locally in our machines.RAG(Retrieval Augmented Generation):The technique of augmenting LLM knowledge with additional data.CAS(Content Aware Storage):A new class of storage that makes data AI ready by extracting the semantic meaning and making it

4、available for consumption via search APIs in a secure,efficient manner.Langchain/Lang-graph:A tool that enables building of LLM models with several capabilities including:prompt-engineering,model selection and building of RAG with external data.Easily integrates with Ollama.Fine-tuning:The Customiza

5、tion of the model and its parameters to optimize the performance for specific tasks or domains4IBM TechXchange|2025 IBM CorporationVector Store5IBM TechXchange|2025 IBM CorporationRetrieval Augmented Generation(RAG)ComponentsIndexingLoad:Load the data using documents loaders.Split:Break the data int

6、o smaller chunks.Store:Store the splits for reference,using VectorStore or Embeddings model Retrieval and GenerationRetrieval:Given the user input,retrieve the relevant split from stored splits.Generate:Generate the answer using a prompt that includes the question and the retriev