Databricks Performance Optimization

Analytics Fundamentals 是一门入门课程，旨在帮助从业者熟悉 Databricks Platform 上的现代分析。本课程涵盖现代分析的五大支柱：湖仓架构、数据转换（重点介绍使用 Databricks SQL 进行基于 SQL 的 ETL）、统一语义、AI 驱动的仪表盘，以及与 Genie 的对话式分析。

注意：Databricks Academy 正在将 Databricks 环境中的课堂课程转为基于 notebook 的形式，并停止在讲座中使用幻灯片。您可以在 Vocareum 实验环境中访问课程 notebook。

Analytics Fundamentals is an introductory course that orients practitioners to modern analytics on the Databricks Platform. This course covers the five pillars of modern analytics: lakehouse architecture, data transformation with an expanded focus on SQL-based ETL using Databricks SQL, unified semantics, AI-powered dashboards, and conversational analytics with Genie.

Note: For SCORM lecture files, please ensure that you close the SCORM window after completing the content. Do not click the ‘Next Lesson’ button, as doing so may prevent the SCORM module from being marked as complete.

In this course, you will learn to design and implement machine learning workflows that scale from a single machine to distributed compute on Databricks. The curriculum is organized into three modules: foundations of model development (when and why to distribute, and how Spark executes work), scalable model development with Spark ML and MLflow, and distributed model training with Ray and PyTorch — closing with a framework for choosing the right tool for a given workload.

Through hands-on demos and labs, you will build distributed Spark ML pipelines using Transformers, Estimators, and the fit/transform paradigm; scale existing pandas code with the Pandas API on Spark; and track, compare, and govern experiments with MLflow 3 and Unity Catalog. You will then train models across a cluster with Ray Train and PyTorch (TorchDistributor and DeepSpeed), and learn to match each workload to the right framework based on data size, model size, hardware, and cost.

The course concludes with a capstone that takes a problem end to end — engineering features, training a model at scale, tracking it with MLflow, and registering it in Unity Catalog with a governance alias. (Production deployment, serving, and monitoring are covered in the separate Advanced MLOps course.)

By the end, you will be able to choose, build, and justify distributed, reproducible, and governed model-development workflows on Databricks.

Note: For SCORM lecture files, please ensure that you close the SCORM window after completing the content. Do not click the ‘Next Lesson’ button, as doing so may prevent the SCORM module from being marked as complete.