|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/master-enterprise-ai-apps/
课程评论：没有评论
**Coursera课程总结：精通企业级AI应用的设计、集成与部署** **课程名称：** Master Designing, Integrating & Deploying Enterprise AI Apps **目标受众：** 机器学习工程师、数据科学家 **课程亮点与学习内容：** 本课程旨在教授企业级数据科学/AI/ML应用的“为什么、是什么、如何做”，重点关注以下方面： * **可扩展的架构设计：** 如何将需求转化为可扩展的架构组件。 * **微服务架构：** 如何运用微服务风格的架构将大型复杂问题拆解为简单易管理的部分。 * **端到端解决方案：** 构建一个完整的、真实的、企业级机器学习解决方案。 * **核心技术栈：** * **异步IO (AsyncIO)：** 学习异步IO的基础知识，以及如何在Python 3中编写I/O密集型应用。 * **NATS：** 了解NATS——一个连接分布式应用的云原生计算基金会开源项目。 * **FlatBuffers：** 掌握FlatBuffers——一种语言独立、紧凑且快速的二进制结构化数据表示语言。 * **Docker & Docker-compose：** 学习在部署和编排应用方面被誉为“黄金标准”的Docker和Docker-compose。 **为何学习这些技术？** * **模型是组件，应用是解决方案：** 统计或深度学习模型只是解决现实世界问题的重要组成部分，而非完整的应用。 * **复杂解决方案的多样性：** 复杂的解决方案通常涉及使用不同语言编写的多个应用程序，并在集群机器上运行。 * **数据科学家/ML工程师的角色拓展：** 您的职责不仅限于模型构建，更需要设计可轻松集成到更大解决方案中的应用，并能通过现代DevOps方法进行部署。 * **提升竞争优势：** 掌握AI应用的集成、可扩展性和可升级性，将为您带来显著的竞争优势。 **教学方式：** 本课程的教学风格结合了三个关键方面，帮助学员全面掌握技术： 1. **直觉（Intuition）：** 通过精美的动画，帮助建立对解决方案的直观理解。 2. **理论（Theory）：** 详细解释技术背后的“为什么”和“是什么”。 3. **代码（Code）：** 提供清晰、愉悦且迭代式的编码实践，从简单的代码示例逐步构建出真实的应用程序。 **课程结构：** * **问题分解与技术选型：** 首先阐述整体目标、相关挑战，并指导如何将复杂问题分解为可管理的组件，从而指导技术选型。 * **技术深入讲解：** 详细介绍所选技术（AsyncIO, NATS, Flatbuffers, Docker）的“为什么、是什么、怎么做”，并配以代码示例。 * **迭代式开发：** 代码示例将以迭代的方式进行，逐步添加功能，最终达到真实应用水平。 * **丰富的学习资源：** 提供所有迭代开发过程中的源代码以及完整的端到端解决方案。
Target Audience Machine Learning Engineers & Data ScientistsWhat is unique about this course & What will you learn?Why What & How of designing, integrating & deploying Enterprise Level Data Science/AI/ML applicationsHow to translate requirements into scalable architectural components?How to break a big complex problem into simple & manageable parts using microservices style architecture?An End-to-End real-world enterprise-level machine learning solutionAsynchronous IO - Foundations & Writing I/O bound applications in python 3NATS - A Cloud Native Computing Foundation open source project to connect distributed applicationsFlatBuffers - A language-independent, compact and fast binary structured data representation languageDocker & Docker-compose - The gold standard in deploying and orchestrating applicationsWhy should you learn all this?A statistical or deep learning model is not an application rather it is an important component of a solution to real-world problems. A sophisticated solution to a complex problem generally consists of multiple applications written using different languages and running on a cluster of machines.Your role as a Data Scientist and Machine Learning engineer is not just limited to a model building or tuning its performance rather it is expected that at the very minimum you will design your applications so that they can easily integrate with other applications of a big solution as well as are easily deployable using modern DevOps methodologies. Mastering how to make AI applications integrate with other applications while ensuring scalability and upgradability will offer you a competitive advantage over others. The good news is that mastering them is not difficult at all!How is this course taught?My teaching style covers 3 key aspects of mastering any technology:IntuitionTheoryCodeFor any solution first I describe the overall goal, its associated challenges, and how to break down a big complex problem into manageable components. This process of simplifying the problems into components will guide you in identifying & selecting the best technology to use. I then explain the why, what & how of the selected technologies (AsyncIO, NATS, Flatbuffers, Docker) with code examples. These code examples start simple and I then iteratively add features to bring them to the level of real-world applications. I have taken immense care in preparing the material that has great animations to help you develop intuition behind the solutions.I have made sure that coding sessions follow an iterative development style and more importantly are clear & delightful. All the source code from the iterative cycles as well as full end to end solution has been provided in the resources.
