Evals for AI Agents with Ravin Kumar (DeepMind) скачать в хорошем качестве

Evals for AI Agents with Ravin Kumar (DeepMind)

This was an impromptu podcast livestream for the Vanishing Gradients podcast: https://vanishinggradients.fireside.fm/ This is the course Ravin and Hugo mentioned: Building AI Applications for Data Scientists and Software Engineers (25% off our final cohort in March for viewers)— https://maven.com/hugo-stefan/buildin... The livestream resolution got worse 1/2 way through but never fear: when we release the podcast, we'll also release the high res recording. Podcast now live here:    • Why Gemini 3 Will Change How You Build AI ...   Check it out! 00:00 Welcome and Introduction 00:57 Guest Introduction: Ravin Kumar from DeepMind 01:25 AI Evaluation and Agent Evaluation 02:34 The Evolution of LLMs and Function Calling 04:21 Tool Calling and Agentic Behavior 08:25 Gemini 3.0 and Advanced AI Capabilities 14:34 Building Effective AI Agents 23:12 Evaluation and User Feedback 30:15 Marmite vs. Vegemite: A Cultural Mix-Up 30:33 Understanding Evaluation Harnesses 31:53 Automating Evaluations for Efficiency 35:11 Complexity in Agent Evaluations 47:23 Multimodal Agentic Systems 56:28 Hosted vs. Open Weight Models 01:03:14 The Future of AI Models and Evaluations