Using a single LLM-powered agent to handle complex workflows sounds appealing — until it starts breaking in unpredictable ways. In this talk, we’ll explore why single-agent systems fail and how multi-agent architectures offer a more scalable and controllable alternative. We’ll introduce the concept of agents, discuss agents as tools and agent-to-agent (A2A) collaboration, and show how workflows can be modeled as coordinated agent interactions. Using Strands as a conceptual framework, we’ll demonstrate how to design and execute an agent-based workflow, culminating in a live demo where multiple agents collaborate to solve a real task. Attendees will leave with a practical understanding of how multi-agent systems work, when they make sense, and how to think about integrating them into real-world backend architectures.

Talk Level:
INTERMEDIATE

Bio:
Kristiyan is a Principal Engineer with over eight years of experience building and evolving large-scale backend systems in fintech and enterprise environments. He has worked extensively on complex system migrations, helping teams modernize platforms while maintaining reliability at scale. His interests span software architecture at multiple levels — from system-wide concerns such as microservices and distributed systems, to service-level design patterns like clean and hexagonal architecture. He is particularly passionate about CI/CD, developer productivity, and using automation to reduce operational and cognitive overhead for engineering teams. An early adopter of LLM-based systems, he focuses on how AI can be introduced into backend architectures in a practical, controlled, and production-ready way. Through his work and talks, he aims to help engineers better understand emerging architectural patterns and apply them with confidence in real-world systems.