Week 0 - Introductory Workshops

Available On-Demand

Theme: Build Your First Agent – Tools, Prompts, and Use Cases

Week 1 - Introductory Workshops | July 16–17

Available On-Demand

Theme: Build Your First Agent – Tools, Prompts, and Use Cases

Week 2 - Intermediate Workshops | July 23–24

Available On-Demand

Theme: Agent Collaboration and Real-World Task Automation

Prashanth Rao, AI Engineer | Kùzu Inc.
Dennis Irorere, Data Engineer | Tripadvisor
‍
Abstract

‍In this hands-on session, we transition from theory to practice, showing how to build intelligent agent workflows powered by Graph RAG and production-ready knowledge graphs. Grounded in practical examples and real tools, this workshop will help you design, construct, and activate agentic systems that are robust, interpretable, and adaptable to complex enterprise environments.

‍Lesson 1: From Business Question to Graph SchemaWe begin by anchoring the session around a real-world business question and discussing how to translate it into a graph schema designed specifically for agentic workflows. You'll learn how to model information in a way that reflects the needs of downstream LLMs and agents, rather than relying on traditional entity-relationship approaches. We’ll also introduce an example data source, and explore how to think critically about the structure, depth, and navigability of the resulting knowledge graph.

‍Lesson 2: From Unstructured Data to Agent-Ready Graphs with BAMLNext, we focus on transforming high-quality unstructured content into structured graph data using BAML, a declarative functional language for building agent-compatible pipelines. You’ll learn how to define reusable extraction logic for entities and relationships, handle ambiguous or nested content, and prepare the data for ingestion. This section emphasizes reliability, modularity, and production-readiness, giving you the tools to automate what is traditionally a manual, high-effort process.
‍
Lesson 3: Agentic Graph RAG in ActionFinally, we’ll walk through a complete Graph RAG loop. You’ll see how an LLM converts a natural language query into Cypher, retrieves data from the Kuzu graph database, and synthesizes a natural language response. We’ll also address what happens when no direct match is found, and how agentic workflows can adapt, recover, and still return useful output.Whether you’re designing internal tools or building external-facing AI agents, this session equips you with the principles and practices needed to take Graph RAG from idea to implementation.

Shane Murray | Field CTO & Head of AI Strategy | Monte Carlo
‍
Abstract

‍Nearly every data organization is under pressure to deploy Agentic AI initiatives, but where do they start? What do data teams actually need to know before they begin?The engineering team at Monte Carlo, a data + AI observability company, not only works with hundreds of enterprise customers to help them build reliable AI products every day, but the team has also developed their own Agentic AI applications within the product.

After a year of deploying GenAI in production, including the world's first-ever Observability Agents, the team has learned 10 key lessons that can serve as guidance for others currently in the process of or planning for their Agentic AI investment. Join Shane Murray, Field CTO & Head of AI at Monte Carlo, as he shares the takeaways his team has learned in the Agentic AI development process.

Lessons include:
‍
- 1: Separate AI research from product research
- 2: Employ observability into model output
- 3: Validate results with data or human interactions
- 4: Divide and aggregate small agentic tasks
- 5: Split tasks horizontally to reduce runtime
- 6: Some tasks will have to wait for better models
- 7: AI solutions can be valuable for fast development
- 8: Establish cooperative workflow between product engineers and data scientists
- 9: Use LLM clone models to keep data secure
- 10: Leverage structured output response formats

Attendees will walk away with a deeper understanding of best practices for Agentic AI product research, validations, task splitting, workflows, security, and perhaps the most difficult, knowing when to be patient.

Session Outline:

Lesson 1: Set the right foundations for the early stages of Agentic AI development. This means establishing the right research workflows and monitoring output quality from the jump. Decide how your team will validate responses.

Lesson 2: Learn how to split Agentic tasks correctly. This means dividing and aggregating small tasks, as well as splitting tasks horizontally to effectively reduce runtime.

Lesson 3. Establish a cooperative workflow between internal teams, from product engineers to data scientists, for continued development and optimization. Learn how to maintain data security with LLM clone models, as well as how to leverage structured output response formats.

I'd focus on lessons around building the Troubleshooting Agent here at Monte Carlo, including:

- The problem definition (and why this problem was suited for AI) - Getting to production / value
- How it's built – a review of the multi-agent architecture built with Claude, LangGraph
- What it does – a demo or overview of the user interaction
- How we evaluate model quality
- Tips for how to get started building agents

Attendees will walk away with a deeper understanding of best practices for product research, validations, task splitting, workflows, security, and perhaps the most difficult, knowing when to be patient when it comes to effective Agentic AI development.

Background Knowledge:

Basic understanding of data engineering, AI development, and data + AI infrastructure.

Week 3 - Advanced Builders Workshops | July 30–31

Live Now! Wed-Thur

Theme: Designing Scalable and Intelligent Agent Systems

David Hughes, Principal Data & AI Solution Architect | Enterprise Knowledge
‍
You've built your Graph RAG system with agentic workflows in the last session—now comes the critical question: How do you know it's actually working? Traditional LLM evaluation metrics fall short when measuring Graph RAG performance in production environments where agents make autonomous decisions based on retrieved knowledge.This advanced session tackles the unique challenges of evaluating agentic Graph RAG systems. Unlike simple question-answering scenarios, agentic workflows require evaluation frameworks that measure not just accuracy, but decision quality, reasoning consistency, and operational reliability under real-world conditions.

‍Lesson 1: What Makes Graph RAG Evaluation DifferentWe'll explore why standard retrieval metrics miss the mark for agentic systems. When agents use Graph RAG to make decisions, you need to evaluate retrieval quality, reasoning coherence, and downstream task performance simultaneously. We'll define evaluation frameworks that capture how well your Graph RAG system enables agents to achieve their goals, focusing on metrics that directly impact business outcomes rather than traditional accuracy measures.

‍Lesson 2: Practical Benchmarking with OPIKYou'll get hands-on experience setting up comprehensive evaluation pipelines using OPIK, an open-source platform designed for production LLM evaluation. We'll demonstrate how to create evaluation datasets that reflect real agentic scenarios, implement automated benchmarking workflows, and interpret results to identify system weaknesses. You'll learn to benchmark not just individual queries, but entire agentic decision sequences that reveal how well your Graph RAG supports agent reasoning.

‍Lesson 3: Production Monitoring and Continuous ImprovementWe'll conclude by looking forward and investigating strategies for ongoing system health monitoring in production environments. You'll learn techniques for detecting performance degradation, handling dynamic data environments, and implementing feedback loops that help your workflows improve over time. We'll cover essential aspects like monitoring data freshness, query performance, and agent decision quality.

We'll also explore multi-agent coordination strategies and how frameworks like Google ADK and A2Acan enhance your agentic workflows to ensure your Graph RAG system maintains reliability as it scales.By the end of this session, you'll have the tools and methodologies to confidently deploy and maintain Graph RAG systems that your agents can rely on for critical decision-making.