Wil discusses his hands-on workshop exploring a fascinating frontier: Graphs as the Memory for AI Agents. Moving beyond simple retrieval, Will explains how we can use graph technology to mirror human cognitive functions, including:

• Episodic Memory: Learning from user interactions.
• Semantic Memory: Building a canonical model of the world.
• Procedural Memory: Unlocking advanced, graph-based reasoning for agents.

Whether you're interested in the latest in GraphRAG, the evolution of Knowledge Graphs in the age of LLMs, or just want to hear about Will’s journey through the startup ecosystem, this conversation is packed with insights that have only become more vital since they were recorded.

In this chat, Paco cuts through the current hype bubble to discuss how graph technologies and entity resolution are solving high-stakes, real-world problems. From interdicting $3 trillion in dark money and human trafficking to fixing systemic data disconnects in government agencies through the NIEM semantic standard, this conversation explores the profound human impact of well-engineered data.

Key Topics:

• Entity Resolution (ER): How connecting data points protects real people in the legal and financial systems.
• Semantic Standards: Using NIEM and SKOS for better context engineering
• Why the biggest bottleneck in AI isn't the math—it's designing UX & visualizations that humans can actually use.
• The AI Bubble: Paco’s perspective on the current industry jitters vs. the tangible tech that is ready for production.

In this episode, Amy Hodler of GraphGeeks is joined by Michelle Yi, co-founder of Generationship, to dive into:

• The "Eval" Crisis: Why 95% of GenAI POCs fail and how rigorous evaluation is the cure.
• Causality & Graphs: Moving beyond "ice cream and shark attacks" to understand the why behind AI predictions.
• Spatial Reasoning: How graphs are becoming essential for robotics and multimodal AI.
• Generationship: Michelle’s mission to fund and support early-stage female+ founders building the future of AI infrastructure.

Connect with Michelle & Generationship:

🔗 https://www.generationship.ai/

Highlights:

• The Spanner Graph Impact: Why Google’s entry into the graph space is a watershed moment for big data and high consistency.
• Human-Centric AI: A look at Knowledge Mapping—helping law enforcement and investigators find absolute truths in unstructured data without relying on AI-generated conclusions.
• No-Code Graphing: Introduction of the Graph Composer, a tool designed to map disparate data sources into a graph model with zero coding.

Learn more about Kineviz: https://www.kineviz.com/

Key Highlights

• Denise argues that the current AI explosion isn't a separate trend but is actually building off previous innovations like graphs that now provide the essential context that AI systems require to function effectively.
• Successful graph-centric companies often see massive valuations because graph technology provides the shortest conceptual path from a business idea to technical implementation, allowing teams to ship faster and align more clearly.
• Denise discusses her new book, Tech Confidential: The Insider’s Playbook for Daring Entrepreneurs. It offers a roadmap for the tech life cycle, covering everything from managing professional egos and building collaborative teams to navigating successful company exits.

https://www.techconfidential.ai/

Key Points:

• Solving the 80% Problem: Despite AI advancements, most effort still goes into data discovery. How to automate extracting metadata like foreign keys from data lakes to build a relationship graph.
• Graph-Based Computation: Wes describes a computational graph where tables are nodes and foreign keys are edges. This turns data preparation into a graph traversal problem, making it faster to roll up data to the right granularity.
• The Return of Facts: As GenAI matures, experts are realizing that text-to-SQL and agents fall short without a robust ground truth. Ontologies and relational metadata are seeing a resurgence as the essential facts AI needs to function.

More data and demo at https://kurve.ai/

In this video:

• Moving beyond text to model audio and video in graphs.
• How graphs extend AI context windows and create long-term procedural memory for agents.
• Better Tooling: Bryce shares how gdotv simplifies the graph stack for developers with better debugging, visualization, and schema scanning.
• Why we still need old-fashioned analytics to understand our data.

Learn more about gdotv: https://gdotv.com/

For over a decade, Stitch Fix has pioneered the blend of human styling and data science. Now, Meg explains how her team is evolving that mission by integrating Knowledge Graphs to better understand their data landscape and inspire the "style journey" for both new and existing clients.

Key Discussion Points:

• Proactive Data Leadership: How Stitch Fix uses graphs to move beyond reactive reporting to being aware of data trends before they surface in a dashboard.
• The Evolution of AI at Stitch Fix: Exploring the intersection of human intuition, generative AI, and graph structures to fulfill client promises.
• ODSC Highlights: Meg’s "aha moment" regarding agent design and the future of AI infrastructure.

Whether you are a data leader or a fashion-tech enthusiast, this conversation offers a unique look at how one of the most data-driven retailers in the world is staying ahead of the curve.

In this interview, they dive into:

• Graph RAG vs. Vector Search: Why similarity search isn't enough for high-stakes enterprise problems and how Graph RAG provides the discernment LLMs lack.
• Neuro-symbolic AI: A look at the locus of reasoning and how combining non-deterministic models with deterministic graph data creates a gray box of explainability.
• The Memory Problem: Why agents need meta-knowledge and a long-term memory that doesn't involve shoving an entire company’s database into a model's context window.
• Neo4j’s Infinigraph: A peek at vertical sharding and how Neo4j is tackling 100-terabyte graphs.

More on Neo4j at https://neo4j.com/

- The AI Evolution: Why ODSC added AI to its name and how the community has shifted from math-first to vibe coding.

- Graphs & AI: The vital role of graph technology in providing memory and context for modern AI systems.

- The Hallway Track: Why in-person community and hands-on learning are the best therapy for tech professionals.

Find an upcoming ODSC event: https://odsc.ai/

The conversation dives into the cutting edge of connected data, moving beyond the initial hype of Retrieval-Augmented Generation (RAG) to explore how graphs are essential for the next generation of AI.

Key Points:

• Multimodal GraphRAG: David explains the shift from simple associative search to a more robust approach that captures signals from images and audio by integrating knowledge management into the graph.
• Agentic Solutions & Observability: A look into the need for engineering rigor in AI agents, focusing on computation graphs, memory inspection, and moving toward deterministic returns using tools like BAML and DSPy.
• The Future of Graph Tech: David discusses the "uncomfortable but exciting" migration from native graph databases to graph-based modeling and querying within vector databases (like LanceDB).

• Zero ETL: Query data lakes and warehouses (SQL, Delta Lake, etc.) as a graph without moving a single byte.
• Scalability: Designed for petabyte-scale analysis in industries like Cybersecurity, Anti-Fraud, and Healthcare.
• Simplify Graph-RAG by turning existing tables into a "knowledge brain" for chatbots.
• Fast Deployment: What used to take six months of data pipelining now takes just weeks via simple schema mapping.

https://www.puppygraph.com/

Amy Hodler (Founder of GraphGeeks) chats with Max Latey (CEO of Pinboard Consulting) to discuss the shift in how organizations are adopting graph technology.

Key Takeaways:

• Graph on Relational: How tools are allowing EAs to sprinkle graph capabilities over existing stacks without heavy ETL.
• GraphRAG & LLMs: Why the push for GenAI is making graph technology a must-have for context-rich AI.
• Data Modeling: Why understanding self-edges is the key to identifying a true network in your data.
• Lowering the Skill Ceiling: How SQL 2023 and GQL are making graph accessible to standard data teams.

Pinboard Consulting specializes in high-impact graph solutions, entity resolution, and architectural strategy. https://www.pinboardconsulting.com/

In this Graph Chat, Bryce Merkel Sasaki sits down with Jessica (Founder of The Ontology Pipeline) to discuss the bridge between Labeled Property Graphs (LPG) and RDF, the rise of neuro-symbolic AI, and why every tech company needs the perspective of a "Chief Librarian."

🔗 Resources:

• Substack: Intentional Arrangement
• Framework: The Ontology Pipeline

Chang shares groundbreaking insights on how agentic retrieval systems are challenging traditional RAG approaches, requiring much higher throughput and iterative search. The conversation highlights the new Lance Format as the multimodal Lakehouse standard optimized for AI data operations.

Most excitingly for the graph community, Chang provides a first introduction on the new open-source project, Lance Graph, which enables storing graph schemas and executing Cypher-like queries directly on Lance tables, integrating vector, tabular, and graph data into a unified format.

Learn why data differentiation is the key to winning in the age of AI agents.

https://lancedb.com/

Hear how the GDC is expanding its focus beyond traditional benchmarks like FinBench and Graphalytics to embrace microbenchmarks, synthetic data generation, and the exciting work being done on LEX schema to potentially unify property graphs and RDF. Learn why many members, including major vendors and researchers, value having a "seat at the table" to shape the future direction of the graph community.

Explore the intersection of data performance, complexity, and the drive for standard graph query languages and schemas essential for emerging AI applications like text-to-graph query.

https://ldbcouncil.org/

Dave Bechberger, currently focusing on MCP servers, agentic memory, and semantic data layers, explains that memory is fundamental because standard LLM calls are atomic and lack recollection of prior interactions. An agent without memory lacks continuity for complex user interactions.

The discussion breaks down three key types of memory and how graphs apply:

• Episodic Memory: Transactional details are directly integrated into the context.
• Short-Term Memory: Session-based interactions that require compaction or summarization.
• Long-Term Memory: For extracting and storing patterns, trends, and preferences across multiple interactions.

GQL Book - https://www.packtpub.com/en-in/product/getting-started-with-the-graph-query-language-gql-9781836204008

Free GQL Playground https://www.ultipa.com/gql-playground

Playbook and resources: https://www.techconfidential.ai/

In this podcast:

• How graphs help make data "AI-ready" by resolving quality issues and providing critical context.
• The difference between what a traditional AI model can do with flat data (what happened) versus what a graph-enhanced model can do (why it happened).
• Why the future of AI might be in smarter, more domain-specific Small Language Models (SLMs) rather than just massive LLMs.
• How Graph RAG can dramatically improve AI outputs.

Follow up with Graphwise on https://graphwise.ai/.

We explore how GraphRAG enhances LLM accuracy by combining the power of knowledge graphs with retrieval-augmented generation, covering everything from vector similarity search to agentic RAG systems. Tomaz also shares his unconventional journey from professional poker player to graph researcher.

Key Topics Covered:

• The importance of incorporating more than just unstructured text
• Advanced retrieval strategies and Text2Cypher generation
• Building knowledge graphs with LLMs and Microsoft's GraphRAG approach

Resources:

📘 Free eBook from Neo4j: https://neo4j.com/essential-graphrag/

Want to take Rocketgraph for a spin? https://rocketgraph.com/free-trial/

Graph practitioners and ontologists will love this:

• Real-world application of graph traversal and pathfinding that's fun!
• Demonstrates how semantic networks can create engaging user experiences beyond traditional data applications
• Explores how people naturally navigate multidimensional concept spaces and associate ideas

Janet relates how she fell in love with what she calls the "beautiful combination of mathematics and art" in laying out graph structures. After pursuing a PhD, she was determined to bring academic research into real-world applications. Despite being warned against studying AI in the 1990s, she followed her passion anyway, and now, thirty years later, we're seeing incredible connections between graphs and AI through LLMs and agentic workflows.

We also get to hear how Janet, as a former competitive roller skater, sees her current product management work as "choreography" - dancing between customer needs, business requirements, and technology constraints while working within deadlines and real-world limitations.

G.V() website: https://gdotv.com/

KGC Conference referenced: https://events.knowledgegraph.tech/

• How RAPIDS and cuGraph are changing graph analytics through GPU acceleration
• Their innovative approach to scaling NetworkX on GPUs without requiring code modifications
• An exploration of GNNs, graph embeddings, and vector search applications

We also looked at current trends in graph technology and their real-world implications. Joe provided pointers to resources (below) for people just getting started as well as his picks for exciting graph-related talk at the upcoming NVIDIA GTC conference.

Getting Started with RAPIDS and cuGraph

https://rapids.ai/

Graph Talks at GTC

https://nvda.ws/4bofZ4e

As a veteran in the graph space, Will shares his personal journey through different graph database architectures and insights on using property graphs and triples for data management.

Additional information:

https://dgraph.io/ https://hypermode.com/ https://github.com/hypermodeinc/modus Recipes on https://github.com/hypermodeinc/modus-recipes

The "the Kleinberg Book" mentioned on Networks, Crowds, and Markets: https://www.cs.cornell.edu/home/kleinber/networks-book/

Hear how this context-rich search goes beyond geospatial to enable users—from government agencies to private enterprises—to find the precise data they need from the vast, ever-growing datasets collected by satellites, drones, and other sources around the globe.

Our guests include Paige Roberts, who’s had various roles in her 25 years in software and is currently the Director of Product Innovation for thatDot. Paige has also co-authored and contributed to several O’Reilly books, including “97 Things Every Data Engineer Should Know” and “Accelerate Machine Learning with a Unified Analytics Architecture.”

We also have our cybersecurity expert, John Cloonan, who has 25 years of experience deploying security controls, combating network attacks, and bringing new cyber solutions to market. John is now the VP of Product at thatDot.

Resources discussed on this podcast:

• Open source streaming graph quine.io/ 
• The blog on scaling - www.thatdot.com/resource-post/scaling-quine-streaming-graph-to-process-1-million-events-second/
• The SANS Institute for Cybersecurity Training and Certification - www.sans.org/  

You’ll learn about the connections between the birth of the World Wide Web and document-based datasets and document stores, such as MongoDB. Amy and Semish also discuss the flexibility of RDF as a reasoning system and ties to the semantic web.

In this quick history of graph databases, you’ll discover the roots of features that we see in modern graph database management systems and gain an appreciation for the collective innovations. 

Semih Salihoğlu Video Series https://www.youtube.com/@KuzuDB

Further Reading on the Early History https://tomandmaria.com/Tom/Writing/VeritableBucketOfFactsSIGMOD.pdf 

2002 Interview with Charlie Bachman https://www.youtube.com/watch?v=iDVsNqFEkB0

https://www.thedatachefs.com/

https://www.knowledgegraph.tech/

Join this GraphGeeks talk with experts Jesús Barrasa and Dave Bechberger to better understand these approaches.