|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/monitoring-and-maintaining-genai-systems/
课程评论：没有评论
课程名称：监控与维护生成式人工智能系统 课程概述：生成式人工智能（GenAI）系统在日常商业操作中越来越普遍且强大，但同时也存在不可预测性、监控复杂和维护困难等挑战。本课程旨在帮助您培养在实际生产环境中监测、评估和维护GenAI系统所需的技能和思维方式。 在本课程中，您将学习如何解读和处理关键性能信号，如延迟、吞吐量、令牌使用率、幻觉率和用户反馈。您将探索如何设计超越传统指标的可观察性层，结合基础设施监控工具（如Prometheus和Grafana）以及以模型为中心的监控工具（如Weights & Biases）。此外，课程还将介绍识别和应对模型漂移、提示失败或质量下降等问题的系统化方法。 您将了解如何将系统健康与业务结果对齐，确保您的GenAI助手在长期内保持相关性、可靠性和可信任性。为使学习更具实践性，课程将通过虚构公司GenPrompt Solutions Inc.开发的GenAI系统InsightBot的故事进行，展示InsightBot如何被监控、审计、更新和优化。 课程结束时，您将掌握实施日志记录和审计跟踪、自动化重训练和部署周期的能力，并利用反馈循环支持持续改进。您还将了解GenAI的MLOps和DevOps最佳实践，以及如何将技术可观察性与伦理人工智能治理和商业战略相结合。 本课程适合数据科学家、机器学习工程师、AI架构师、DevOps专业人员和与GenAI系统相关的技术负责人。无需具备监控工具的先前经验，课程会逐步引导您。如果您准备好从构建GenAI系统转向自信和负责任地运行它们，这门课程是您下一步的理想选择。
Generative AI systems are powerful, dynamic, and increasingly integrated into everyday business operations - but they are also unpredictable, complex to monitor, and difficult to maintain over time. This course is designed to help you build the skills and mindset needed to monitor, evaluate, and maintain GenAI systems in live production environments.In this course, you'll learn how to interpret and act on key performance signals such as latency, throughput, token usage, hallucination rate, and user feedback. You'll explore how to design observability layers that go beyond traditional metrics - integrating both infrastructure-level monitoring (with tools like Prometheus and Grafana) and model-centric monitoring (with Weights & Biases).We'll also walk through structured approaches to identifying and responding to issues like model drift, prompt failure, or quality degradation. You'll understand how to align system health with business outcomes, and how to ensure your GenAI assistant stays relevant, reliable, and trustworthy over time.To make the learning practical and grounded, you'll follow the story of InsightBot, a GenAI system developed by a fictional company - GenPrompt Solutions Inc. You'll see how InsightBot is monitored, audited, updated, and optimized as part of an ongoing system lifecycle.By the end of this course, you'll understand how to implement logging and audit trails, automate retraining and deployment cycles, and use feedback loops to support continuous improvement. You'll also gain awareness of best practices in MLOps and DevOps for GenAI, and how to connect technical observability with ethical AI governance and business strategy.This course is ideal for data scientists, machine learning engineers, AI architects, DevOps professionals, and technical leads working with GenAI systems. No prior experience with monitoring tools is required - the course will guide you step by step.If you're ready to move from building GenAI systems to running them confidently and responsibly, this course is your next step.
