Safe and Sound AI

As AI evolves from single agents into complex, multi-agent systems, the challenge of monitoring and trusting these autonomous collaborators grows. Traditional monitoring tools fall short, unable to interpret the dynamic, unpredictable nature of AI decision-making.  This discussion explores "Agentic Observability," a new approach built to provide deep visibility into an agent's entire operational lifecycle: thought, action, execution, reflection, and alignment.  By understanding the complete reasoning process, this paradigm moves beyond simple monitoring to become a necessary control layer, providing the transparency and trust required to unlock the true potential of sophisticated AI systems. Learn more:  https://www.fiddler.ai/blog/agentic-observability-development https://www.fiddler.ai/blog/anatomy-ai-agent

In this episode of Safe and Sound AI, we dive into the challenge of moving AI agents from impressive demos to robust, production-ready systems. We break down the principles of Agentic Observability, explaining how this essential "blueprint" provides the clarity needed to overcome the "black box" problem during both development and production.  Learn practical methods for monitoring key signals like tool usage and planning, discover how to diagnose the root causes of agent failures, and explore strategies for ensuring your agent delivers real-world value.

In this episode of Safe and Sound AI, we dive into the challenge of drift in machine learning models. We break down the key differences between concept and data drift (including feature and label drift), explaining how each affects ML model performance over time. Learn practical detection methods using statistical tools, discover how to identify root causes, and explore strategies for maintaining model accuracy. Read the article by Fiddler AI and explore additional resources on how AI Observability can help build trust into LLMs and ML models.

In this episode, we discuss the new integration between Fiddler Guardrails with NVIDIA Nemo Guardrails, pairing the industry's fastest guardrails with your secure environment. We explore the setup process, practical implications, and the role of the Fiddler Trust Service in providing guardrails, monitoring and custom metrics. Plus, we highlight the free trial opportunity to experience Fiddler Guardrails firsthand.  Read the article to learn more, or sign up for the Fiddler Guardrails free trial to test the integration for yourself.

In this episode, we explore how Fiddler Guardrails helps organizations keep large language models (LLMs) on track by moderating prompts and responses before they can cause damage. We break down its industry best latency, secure deployment options, and how it works with Fiddler’s AI observability platform to provide the visibility and control to adapt to evolving threats. Read the article to learn more about how Fiddler Guardrails can help safeguard your LLM Applications.

In this episode, we explore two key approaches for monitoring AI models: metrics and inference observation. We break down their trade-offs and provide real-world examples from various industries to illustrate the advantages of each model monitoring strategy for driving responsible AI development.  Read the article by Fiddler AI and explore additional resources for more information on how AI observability can help developers build trust into AI services.

In this episode, we discuss how to monitor the performance of Large Language Models (LLMs) in production environments. We explore common enterprise approaches to LLM deployment and evaluate the importance of monitoring for LLM quality or the quality of LLM responses over time. We discuss strategies for "drift monitoring" — tracking changes in both input prompts and output responses — allowing for proactive troubleshooting and improvement via techniques like fine-tuning or augmenting data sources. Read the article by Fiddler AI and explore additional resources on how AI observability can help developers build trust into AI services.