Google Cloud Professional Machine Learning Engineer Crash Course
Published by O'Reilly Media, Inc.
Master AutoML, BigQuery, ML pipelines, and more
Course outcomes
- Know what topics are on the exam and how to best study for each of them
- Explore feature engineering, Google Cloud tools, model monitoring, and pipeline systems
- Follow a clear roadmap to getting certified as a Professional Google Cloud Machine Learning Engineer
Course description
Join expert Richard Bryant to delve into the dynamic world of Google Cloud platform and elevate your expertise by navigating through essential tools such as AutoML, BigQuery, Dataflow, and Dataproc. You’ll grasp orchestration frameworks such as Kubeflow and Cloud Composer and explore various modeling techniques and serving models. You’ll work through hands-on exercises to unravel the intricacies of these topics, gain the prowess to discern optimal strategies for diverse scenarios, and emerge ready to apply these skills in real-world ML scenarios, pass the exam, and enhance your career opportunities.
What you’ll learn and how you can apply it
- Perform feature engineering and feature crosses at the right time
- Perform data processing and ML in GCP using the appropriate tool for the job
- Apply an appropriate pipeline orchestration framework for a problem
- Instantly recognize problems such as the bias-variance trade-off and data leakage when training and serving models
- Serve models using Google Cloud appropriately using batch or online prediction, with proper considerations for issues like security and data drift/skew
This live event is for you because...
- You’re a data scientist, data engineer, software/cloud developer, or machine learning engineer who’s looking to move into a senior or principal role.
Prerequisites
- Access to a Google Cloud environment
- Some familiarity with Google Cloud or another cloud platform (preferred, but not required)
- Exposure to mathematics at a college level
- Some familiarity with statistics and machine learning fundamentals
- Basic knowledge of SQL and Python
Recommended follow-up:
- Read Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow (book)
- Read Machine Learning Engineering with Python (book)
- Read Designing Machine Learning Systems (book)
Schedule
The time frames are only estimates and may vary according to how the class is progressing.
Overview of GCP tools (65 minutes)
- Presentation: BigQuery, Cloud SQL, and Bigtable; Vertex AI; Dataproc and Dataflow; AutoML versus prerained APIs; what these tools are used for and key differences between them
- Hands-on exercises: Explore tools
- Q&A
- Break
Feature engineering (20 minutes)
- Presentation: Motivation and best practices; intro to feature crosses; Vertex AI Feature Store and Memorystore; when to perform feature engineering and how Google Cloud facilitates it
Model building (70 minutes)
- Presentation: Capabilities of Vertex AI; review of ML concepts; modeling techniques and their differences (CNNs, RNNs, federated learning, recommendation engines); situations that call for each of these models; model explainability using Vertex AI
- Hands-on exercise: Build a model using Vertex AI
- Q&A
- Break
ML infrastructure I (25 minutes)
- Presentation: TensorFlow fundamentals and common errors; distribution strategies and their trade-offs (mirrored, multiworker, TPU, parameter server); hardware differences; TensorFlow Enterprise versus TFX Break
ML infrastructure II (50 minutes)
- Presentation: Model monitoring (detection of drift, skew, and data leakage); online versus batch serving; orchestration strategies and differences between frameworks (Cloud Composer, Kubeflow, Apache Airflow, Vertex AI Pipelines)
- Hands-on exercise: Explore a full machine learning pipeline with monitoring
Wrap-up and Q&A (10 minutes)
Your Instructor
Richard Bryant
Richard Bryant started his career as a data scientist in the healthcare industry in 2016. He loves using data to solve complex problems with the goals of improving patient care and helping institutions make more data-driven decisions to save cash. Previously, he worked as a graduate student instructor and would later infuse his passion for teaching with his love for data science with the creation of the YouTube channel RichardOnData. Today his channel has 24,000 YouTube subscribers and has been viewed over 850,000 times. Richard earned a master’s degree in statistics from the University of Michigan. He loves exercising, hiking, chess, and his cat Iris.