Automating Workflows with a Deterministic Network of Modular Agents

Learn to design and implement production-grade AI agent systems through a deterministic network architecture in this 32-minute conference talk from MLOps World. Explore how to move beyond brittle autonomous agents or overly scripted systems by structuring workflows as directed acyclic graphs (DAGs) where each node represents a narrow, task-specific agent. Discover the balance between structure and flexibility that enables controlled decision-making while maintaining system reliability, with clear execution paths that support debugging and retries. Examine a real-world case study of an agent system built for private equity and venture capital workflows that continuously ingests pitch decks, performs market research, and generates interactive reports, demonstrating significant reduction in manual effort while preserving quality and interpretability. Gain practical insights into making agent systems modular, reliable, and aligned with actual business workflows, moving beyond simple prompt-based approaches to build scalable systems that automate complex operational tasks in production environments.