Beyond Prompts: Building GenAI Systems with Knowledge Graphs

Most GenAI applications start with prompts or simple workflows, but these approaches often break down when applied to real-world problems. As systems scale, issues like lack of context, unreliable reasoning, and limited control become increasingly visible.

In this session, we’ll explore how knowledge graphs bring structure, grounding, and consistency to GenAI systems. Through an interactive walkthrough and a live demo, you’ll see how combining LLMs with structured knowledge helps move from basic prototypes to systems that are more reliable and production-ready .

This is a practically driven session designed to help practitioners understand how to build structured, reliable, and scalable GenAI systems beyond prompt-based approaches .

Key Takeaways:

• Understand the limitations of prompt-based GenAI – why simple workflows fail in real-world scenarios and where gaps emerge
• Learn how knowledge graphs enhance GenAI systems – how structured knowledge improves context, grounding, and reasoning
• Explore how GenAI systems evolve beyond prompts – moving from single responses to multi-step, structured workflows
• See a practical implementation – a live demo showing how to integrate knowledge graphs into a GenAI pipeline
• Build a stronger system design intuition – how to think about designing GenAI systems that are more reliable and controllable

Joseph is an AI Research Engineer at Domyn, focused on building applied GenAI systems using knowledge graphs, agent-based workflows, and structured reasoning.

He has worked across domains including healthcare and fintech, and brings experience in developing AI solutions for complex, real-world use cases. Previously, he was a founder, giving him practical insight into building and scaling products.You can connect with him on LinkedIn .