DGX Spark Live: Process Text for GraphRAG With Up to 70B LLM
- Date
- 2025-11-21
- Host
- NVIDIA
About this event
Modern GraphRAG pipelines rise or fall on one thing: how well you turn messy text into structure a model can actually use. DGX Spark Live: Process Text for GraphRAG With Up to 70B LLM is a focused, in-person session for people who want to understand that step in practical terms, not as vague architecture diagrams. If you care about retrieval quality, entity extraction, knowledge graphs, or what large models can realistically do inside a GraphRAG workflow, this meetup is built for you. About the Event This event centers on a very specific and timely challenge in AI systems: processing text for GraphRAG using large language models at substantial scale. The title signals a hands-on, technically serious conversation around working with models up to 70B parameters and applying them to the text-to-graph layer that powers better retrieval, reasoning, and downstream responses. Rather than treating GraphRAG as a buzzword, the session is positioned around the real pipeline question many builders face: how do you go from raw documents to graph-structured knowledge that an LLM can actually retrieve and reason over effectively? Expect a discussion grounded in implementation choices, tradeoffs, and workflow design. Because this is a live, in-person meetup, the format also matters. You are not just watching passively; you are stepping into a room with other people actively thinking about LLM infrastructure, retrieval systems, applied AI, and where graph-based methods fit into modern stacks. That makes this useful both as a learning session and as a place to compare approaches with peers. What to Expect The core of the event will likely revolve around the text-processing layer of a GraphRAG system. That means attention on tasks such as breaking down unstructured content, identifying entities and relationships, organizing information into graph-friendly representations, and understanding how larger models can support those steps. You should expect a session that helps connect the dots between LLMs, retrieval design, and graph-based knowledge representation. Even if your background is stronger in one area than the others, this is the kind of topic that rewards seeing the full pipeline end to end. Likely areas of focus include: How text is transformed from raw source material into structured graph components Where large models help most in GraphRAG workflows, especially at the extraction and enrichment stages Tradeoffs of model scale, including when using larger models may improve quality or flexibility System design considerations for teams building retrieval pipelines that need more than standard vector search Live discussion and networking with others working across AI engineering, experimentation, and applied use cases Since the event is tagged with community, networking, and meetup, expect interaction beyond the main session. That could mean informal Q&A, post-talk conversations, and practical exchanges with attendees who are building, evaluating, or exploring similar systems. Why Attend If you are building with LLMs right now, GraphRAG is increasingly relevant because many real-world applications need more structure than plain semantic search can provide. Better graph construction can improve retrieval quality, contextual grounding, and the model’s ability to navigate relationships across documents, entities, and concepts. This session speaks directly to that need. The specific emphasis on processing text is especially valuable. Many AI events stay at the strategy level; this topic gets closer to the part that determines whether a GraphRAG system is actually useful in practice. Understanding how text becomes a graph is essential if you want to improve answer quality, reduce retrieval noise, or support more complex reasoning over your data. You should attend if you want clearer intuition around questions like: What role should a large model play in extraction versus retrieval? When does graph structure add value beyond embeddings alone? What gets harder as you move toward larger models and richer pipelines? How do you think about quality, consistency, and usefulness in graph-oriented text processing? There is also a strong practical upside to being in the room. In-person events create space for the kinds of detailed conversations that rarely happen in comment threads or polished conference talks. If you want to compare tooling choices, ask direct technical questions, or meet others thinking seriously about advanced retrieval systems, that alone can make the time worthwhile. Practical Details Location: In person Date: Friday, November 21 Time: 11:00 AM PST Because this is an in-person event, plan for a more interactive experience than a webinar or recorded session. You will be able to listen, ask questions, and connect with other attendees face to face, which is particularly helpful for technical topics where follow-up questions often matter as much as the main presentation. This event is a strong fit for people in AI, LLM, and applied research communities who want substance over hype. The title is specific, the topic is current, and the format gives you a chance to engage directly with both the ideas and the people around them. If GraphRAG is on your roadmap, already in your stack, or still something you are actively evaluating, this meetup offers a focused way to sharpen your understanding of how large models can be used to process text into more useful knowledge structures.
Who should attend
If you are working on retrieval, knowledge-rich AI systems, or practical LLM applications, this meetup should feel highly relevant. - You are **building or evaluating RAG systems** and want to understand when graph-based retrieval adds meaningful value beyond a standard vector pipeline. - You work with **LLMs in production or prototyping** and want a clearer view of how larger models can help with text extraction, enrichment, and structuring for downstream retrieval. - You are an **AI engineer, ML practitioner, data scientist, or technical founder** looking for concrete ideas on turning unstructured documents into something more queryable and useful. - You are exploring **GraphRAG, knowledge graphs, entity extraction, or document understanding** and want to connect those concepts to real implementation decisions. - You learn best by being **in the room with other sharp practitioners**, asking questions live, and comparing approaches with people solving similar problems. - You are part of the broader **AI and LLM community** and want a meetup that is focused enough to teach you something specific while still leaving room for good technical conversation and networking.