MongoDB at
Ai4 2025

This hands-on workshop equips you with the skills to create intelligent agents that seamlessly process both visual and textual information, a critical capability for next-generation applications.

Discover why multimodal agents represent a quantum leap beyond simple prompting and RAG systems, with concrete examples of how they solve previously intractable problems in document processing and visual reasoning. Through practical demonstration, we deconstruct the four pillars of effective agent architecture: perception, planning/reasoning, tools, and memory.

The workshop reveals breakthrough approaches to mixed-modality document handling, showcasing how modern VLM-based embedding models overcome traditional limitations while dramatically simplifying implementation pipelines.

You'll build a multimodal agent using Python, Gemini 2.5, and MongoDB that can: - Answer complex questions about documents containing interleaved text and visuals - Explain charts and diagrams with human-like reasoning - Maintain conversational context across multiple interactions

This isn't theoretical, you'll leave with working code you can immediately adapt for your projects, saving weeks of development time. Whether you're building customer-facing applications or internal tools, these techniques provide a competitive edge in the rapidly evolving AI landscape.

Modeling & Sizing your AI Workload

Tuesday, August 12 | 11:00am | MongoDB Booth

15-minute Booth Session

SPEAKER(S): Sigfrido Narvaez

The presentation includes insights from the book Build AI Intensive Python Applications for which Sig Narvaez is a co-author.

• The talk focuses on MongoDB schema design for AI-driven applications.
• It covers optimizing data structures for various search strategies.
• Strategies for scaling deployments from single to multi-cloud regions are explored.
• Practical considerations for sizing and cost management are also addressed.

Don't Say NoSQL to Me

Tuesday, August 12 | 12:00pm | MongoDB Booth

15-minute Booth Session

SPEAKER(S): Tom Cordova-Caddes

This lightning session in the MongoDB booth will cover what makes MongoDB different and why that makes it the perfect fit for your AI workloads.

Building Scalable Memory into your Agentic Apps

Tuesday, August 12 | 1:00pm | MongoDB Booth

20-minute Booth Session

SPEAKER(S): Henry Weller

AI builders know that integrating memory easily into your AI applications with MongoDB is only the beginning; once your app starts to see significant data and query volumes, the real fun begins. This session will explore different considerations for building scalable vector search into your app, considering recall, cost, and latency/throughput as primary considerations.

Though technical, this talk will cater to those more familiar with embedding models and LLMs, even if they are less familiar with databases, search systems.

Powering AI Agents with Your Data: MongoDB + deepset & Haystack

Wednesday, August 13 | 11:00am | MongoDB Booth

20-minute Booth Session

SPEAKER(S): Jay Wilder, VP of Growth at Deepset AI

AI agents are rapidly becoming the interface between people and systems—and systems themselves––but creating agents that are truly useful in enterprise environments requires deep domain context, reliable data access, and flexible orchestration.

In this talk, we’ll show how MongoDB and the open-source Haystack framework and tools by deepset make it easy to build custom AI agents that reason, retrieve, and act—grounded in your organization’s structured and unstructured data.

Through the integration of MongoDB Atlas and deepset, developers can rapidly spin up agents that connect to live business data, access relevant documents, and interact through natural language. With Haystack as the foundation, these agents can be extended with tools, memory, and logic tailored to real-world workflows—from customer support and sales ops to document analysis and internal knowledge assistants.

Whether you're building your first agent or scaling to a fleet, this session will show how to go from raw data to intelligent, production-ready agents—faster and without the usual complexity of ML ops or custom infrastructure.

Chat, Search, Resolve: The Future of Network Monitoring using Autonomous AI

Wednesday, August 13 | 12:00pm | MongoDB Booth

15-minute Booth Session

SPEAKER(S): Sourav Biswas

This presentation and demo will showcase how MongoDB MCP servers can be integrated with AI agents and chatbots. These intelligent systems will act autonomously to observe systems in real-time, detect anomalies, and derive actionable insights. We'll also demonstrate how to build a custom tool for vector search, a capability not natively supported by the current MCP server.

GraphRAG: Integrating LLMs with Knowledge Graphs

Wednesday, August 13 | 1:30pm | MongoDB Booth

15-minute Booth Session

SPEAKER(S): Thibaut Gourdel

While traditional RAG is effective, it can struggle with complex relationships and reasoning across large knowledge bases. GraphRAG, an advanced variant, addresses these challenges by leveraging knowledge graphs to enable deeper understanding and improved response accuracy.

Learn how LLMs extract key entities and relationships from your data to construct a graph structure, and how the system uses graph traversal to find related entities and enrich prompts. Stay for a live demo showcasing these concepts in action.