Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn how to build scalable, enterprise-ready AI agents using data flywheels through insights from NVIDIA's real-world deployments in this 20-minute conference talk. Discover how Santiago from NVIDIA frames AI agents as digital employees and explores the architecture behind NV Infobot, including its various use cases for chatbots, copilots, and 'talk-to-your-data' solutions. Understand the MAPE (Monitor, Analyze, Plan, Execute) framework for creating self-regulating agents and examine the challenges of collecting user feedback effectively. Explore root cause analysis techniques using LLM assistance, learn about key error types and their prioritization strategies, and see how to construct data flywheels that drive continuous improvement of large language models through user feedback. Examine NVIDIA's microservices approach to agent development and follow the developer workflow for fine-tuning models. Study two practical experiments: router optimization and query rephrasal improvement, demonstrating how these architectural patterns maintain accuracy, scalability, and adaptability in fast-paced business environments. Access the W&B Blueprint for deploying your own data flywheel and gain insights into error prioritization and Nemo tools through the Q&A session.
