Building Trust in LLM Systems with Agentic Observability with Krisha Gade from Fiddler AI

As LLM applications evolve into complex, multi-agent systems, traditional monitoring tools fall short in ensuring reliability and trust. In this session, Fiddler will introduce Agentic Observability—a new approach to debugging and monitoring production LLMs at scale. Drawing from real-world deployments across enterprise and federal teams, the talk will cover techniques for tracing agent reasoning, evaluating output quality with Fast Trust Models, and detecting behavioral shifts using statistical and embedding-based methods.

The session also explores how to simulate and test agent workflows efficiently within the constraints of modern AI systems, offering practical insights for engineering trustworthy and scalable GenAI applications.

Register for @Scale: AI & Data on June 25 here: https://atscaleconference.com/events/scale-data-ai-infra/

Видео Building Trust in LLM Systems with Agentic Observability with Krisha Gade from Fiddler AI канала @Scale