|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/building-and-evaluating-llm-powered-apps-on-aws/
课程评论：没有评论
课程名称：在AWS上构建与评估大语言模型（LLM）驱动的应用程序 课程概述：本课程《在AWS上构建与评估LLM驱动的应用程序》提供了一个全面而实用的学习之旅，带您深入了解大型语言模型（LLMs）及其在亚马逊云服务（AWS）上的应用开发，特别关注于Amazon Bedrock。您将首先了解Amazon Bedrock的能力，这是一种完全托管的服务，提供无服务器访问，能够接触到来自顶尖AI提供商（如Anthropic、AI21 Labs、Cohere和亚马逊自己的Titan模型）的多种高性能基础模型（FMs）。我们将阐释为什么Bedrock对开发者而言是一个颠覆性的工具，它简化了模型托管和基础设施管理的复杂性。 课程的核心内容为实践操作。您将学习如何为LLM开发设置安全高效的AWS环境，包括IAM角色和权限的配置，并熟练掌握AWS命令行界面（CLI）和Boto3 SDK以进行与Bedrock的程序化交互。这一基础知识将使您能够直接与各种LLM进行交互，试验不同的模型参数（如温度和top-p），并利用聊天游乐场进行快速原型设计和提示工程。 课程中有相当一部分专注于构建复杂的LLM应用程序。您将深入学习如何使用AWS Bedrock Agents构建智能代理，设计其工作流程，整合自定义工具通过AWS Lambda函数来扩展其功能（例如获取实时数据或与外部API交互），以及处理复杂的多步骤任务。您还将掌握增强生成（RAG）的艺术，这是一种通过使用您自己的专有数据来提高LLM响应质量的强大技术。这包括将文档嵌入并索引到知识库中，执行向量搜索，以及增强LLM提示以生成上下文丰富且准确的答案。 课程的重要环节还包括对LLM应用程序的评估技能的学习。我们将讨论各种评估技术，包括使用Amazon Bedrock的“LLM作为评判者”功能，以及进行输出比较和评分的方法。您将学习如何衡量关键指标，例如响应质量、事实正确性（最小化幻觉现象）和与用户查询相关性，确保您的应用程序不仅是功能性，还在实际场景中表现良好、可靠。 通过本课程的学习，您将掌握设计、开发、部署和严格评估自己智能化、具备生产水平及成本意识的LLM驱动应用程序的实践技能和自信，无论是用于聊天机器人、知识助手还是新颖的生成AI解决方案。
This course, "Building and Evaluating LM-Powered Apps on AWS," offers a comprehensive and practical journey into the world of Large Language Models (LLMs) and their application development on the Amazon Web Services (AWS) cloud, with a strong focus on Amazon Bedrock.You'll begin by gaining a solid understanding of Amazon Bedrock's capabilities as a fully managed service that provides serverless access to a diverse range of high-performing Foundation Models (FMs) from leading AI providers like Anthropic, AI21 Labs, Cohere, and Amazon's own Titan models. We'll demystify why Bedrock is a game-changer for developers, abstracting away the complexities of model hosting and infrastructure management.The core of the course is intensely hands-on. You'll learn to set up a secure and efficient AWS environment for LLM development, including the configuration of IAM roles and permissions, and mastering the use of the AWS Command Line Interface (CLI) and Boto3 SDK for programmatic interaction with Bedrock. This foundational knowledge will empower you to interact directly with various LLMs, experiment with different model parameters (like temperature and top-p), and utilize the Chat Playground for rapid prototyping and prompt engineering.A significant portion of the course is dedicated to building sophisticated LLM applications. You'll dive deep into building intelligent agents using AWS Bedrock Agents, learning how to design their workflows, integrate custom tools via AWS Lambda functions to extend their capabilities (e.g., fetching real-time data or interacting with external APIs), and handle complex, multi-step tasks. You'll also master the art of Retrieval-Augmented Generation (RAG), a powerful technique to enhance LLM responses by grounding them with your own proprietary data. This involves practical steps like embedding and indexing documents in a knowledge base, performing vector searches, and augmenting LLM prompts to generate contextually rich and accurate answers.Crucially, the course doesn't stop at building. You'll learn the vital skill of evaluating your LLM applications. We'll cover various evaluation techniques, including the use of Amazon Bedrock's "LLM-as-a-judge" feature, and methods for running comparisons and scoring outputs. You'll learn to measure key metrics such as response quality, factual correctness (minimizing hallucinations), and relevance to user queries, ensuring your applications are not only functional but also performant and reliable in real-world scenarios.By the conclusion of this course, you will possess the practical skills and confidence to design, develop, deploy, and rigorously evaluate your own intelligent, production-ready, and cost-aware LLM-powered applications on Amazon Bedrock, whether for chatbots, knowledge assistants, or novel generative AI solutions.
