Balancing Autonomy and Determinism in Agentic Systems

LLM-powered agents promise a future of intelligent automation, but their autonomy often comes at the cost of reliability. How do we build agents that are both smart and dependable?

This session explores a key question in AI development: how do we build autonomous systems that are creative yet also reliable? Using a practical agent that turns user content into presentations, we'll break down the core design decisions that lead to consistent and resilient AI. We’ll cover essential techniques, including proper tool design, Pydantic validation, strategic prompt engineering, and more.

About the Speaker
Hansin Patwa is a Master's student in Computer Science at UC San Diego with a passion for building practical applications that connect LLM reasoning with data and tools. His work focuses on agentic AI systems, with hands-on experience in frameworks like LangGraph. For his recent "slide Agent" project, he engineered an agent that transforms documents into presentation-ready Google Slides. When not experimenting with the latest in AI, he can be found cooking food or playing video games.