Agentic AI Architectures: From Single-Agent to Multi-Agent (with a Practical Case Study)

AI systems are rapidly evolving beyond simple prompt-response interactions to become goal-driven, decision-making systems. However, many LLM-based applications struggle when applied to real-world scenarios due to lack of planning, memory, and structured execution.

In this session, we will explore how agentic architectures introduce planning, memory, and tool integration to build more reliable AI systems. We’ll compare single-agent vs multi-agent design patterns and walk through a practical case study where multiple agents, planner, executor, and reflector collaborate to solve a task end-to-end.

This is an intuition-first, practically grounded session designed to help you understand how agentic patterns translate into real-world AI systems.

Key Takeaways:

• Evolution of AI Systems – from prompt-based interactions to goal-driven agentic workflows
• Agentic Architectures – understanding planning, memory, and tool usage in AI systems
• Single vs Multi-Agent Systems – key design patterns and when to use each
• Practical Case Study – how planner, executor, and reflector agents collaborate AI systems
• System Design Intuition – building reliable, scalable agent-based AI applications

Lakshmi Devi Prakash is Vice President – AI/ML Lead at JPMorgan Chase, based in Bengaluru, with over a decade of experience in Machine Learning, NLP, and Generative AI.

She specializes in building production-ready AI systems, including enterprise solutions, agent-based architectures, and multimodal applications. She is currently pursuing a PhD in AI, has published research papers, authored books on Generative AI, and holds a patent in AI-driven automation.