Scaling Configuration-Driven Multi-Agent Networks from Prototype to Production
Details
RSVP here - https://streamyard.com/watch/a8MAcYY7NShj
Speaker: Daniel Fink, Associate Vice President — Platform Engineering at Cognizant
Many enterprise AI systems still rely on a single large language model to do everything such as retrieve, reason, act, and respond. But real-world problems aren’t solved by solo agents, they are solved by teams.
In this session, we will explore a new design paradigm for building collaborative, multi-agent systems that mirror how real teams work: distributed, specialized, and task-focused. We will show this shift unlocks new capabilities that single-agent LLMs can’t handle, like coordinating across tools, managing context, enforcing data boundaries, and scaling decision-making workflows.
This session will also show how to prototype these agentic systems using a configuration-driven approach that lets you define agents, roles, tools, and communication patterns without writing orchestration code. This allows teams to quickly test, adapt, and scale agent-based architectures across real enterprise scenarios.
We’ll walk through applied examples that blend generative AI with traditional software capabilities, such as querying APIs, triggering actions, handling private data securely, and collaborating across agents with role-specific logic and fallback strategies.
The focus of this session will reframe how multi-agent AI is perceived and and show attendees how to start building today. Attendees will be able to translate real-world use cases (RAG, agentic automation, intelligent assistants) into scalable, testable multi-agent networks.
## Useful Links
- Free access to more talks/trainings: Ai+ Training platform
- ODSC blog: https://opendatascience.com/
- Slack Channel: https://hubs.li/Q038cQBy0
- Code of conduct: https://odsc.ai/code-of-conduct/