Google AI Course

Lesson 1 • Core AI Subfields Overview

Maps machine learning, NLP, computer vision, and robotics as distinct disciplines. Helps learners navigate which Google tools belong to which subfield.

Lesson 2 • How Machines Learn from Data

Explains supervised, unsupervised, and reinforcement learning paradigms. Connects learning types to practical Google AI product categories.

Lesson 3 • What AI Is and Is Not

Clarifies common misconceptions and establishes precise definitions of AI. Sets the conceptual baseline for all subsequent technical content.

Lesson 4 • History and Evolution of AI

Traces AI from symbolic logic to modern deep learning. Provides context for why current Google AI tools are designed as they are.

Lesson 1 • Cloud AI and Machine Learning Platform

Introduces Vertex AI as the unified ML platform and its core services. Connects cloud infrastructure to practical model training and deployment workflows.

Lesson 2 • Pre-built AI APIs for Developers

Covers Vision, Speech, Translation, and Natural Language APIs as ready-to-use services. Shows how pre-built APIs accelerate development without custom training.

Lesson 3 • AI in Google Workspace and Consumer Products

Explores Gemini integrations in Docs, Sheets, Gmail, and Search. Demonstrates how embedded AI boosts everyday productivity for non-technical users.

Lesson 4 • Generative AI Products and APIs

Surveys Gemini models, Imagen, and related generative APIs available to developers. Positions each product within the broader Google AI ecosystem.

Lesson 1 • Data Collection and Ingestion

Covers APIs, web scraping, streaming, and batch ingestion pipelines. Connects data acquisition methods to Google Cloud storage and BigQuery.

Lesson 2 • Dataset Splitting and Versioning

Teaches train, validation, and test splits and dataset version control practices. Prevents data leakage and ensures reproducible experiments.

Lesson 3 • Data Labeling and Annotation

Explains manual, semi-automated, and automated labeling strategies using Vertex AI Data Labeling. Quality labels are the foundation of supervised learning performance.

Lesson 4 • Data Types and Sources

Distinguishes structured, unstructured, and semi-structured data and their AI use cases. Establishes data literacy as the prerequisite for all modeling work.

Lesson 1 • Model Evaluation and Selection

Interprets confusion matrices, ROC curves, and feature importance outputs from Vertex AI. Guides the decision to deploy, retrain, or discard a model.

Lesson 2 • AutoML for Tabular and Image Data

Trains no-code models on tabular, image, text, and video datasets using AutoML. Demonstrates how non-ML-experts can produce production-quality models.

Lesson 3 • Custom Model Training on Vertex AI

Runs custom TensorFlow and PyTorch training jobs using Vertex AI Training. Covers job configuration, hardware selection, and distributed training basics.

Lesson 4 • Vertex AI Workbench and Notebooks

Sets up managed notebooks for exploratory data analysis and model prototyping. Provides the hands-on environment used throughout all training exercises.

Lesson 1 • CI/CD for ML Pipelines

Automates model retraining and redeployment using Vertex AI Pipelines and Cloud Build. Applies software engineering best practices to the ML lifecycle.

Lesson 2 • Batch Prediction Pipelines

Runs large-scale offline inference jobs using Vertex AI Batch Prediction. Optimizes cost and throughput for non-latency-sensitive workloads.

Lesson 3 • Model Monitoring and Drift Detection

Configures Vertex AI Model Monitoring to detect feature and prediction drift. Ensures deployed models maintain accuracy as real-world data distributions shift.

Lesson 4 • Online Prediction Endpoints

Deploys models to Vertex AI endpoints for low-latency real-time inference. Covers traffic splitting, scaling policies, and endpoint health checks.

Lesson 1 • Bias Mitigation Strategies

Implements pre-processing, in-processing, and post-processing debiasing techniques. Reduces discriminatory outcomes while preserving model utility.

Lesson 2 • Explainability and Transparency

Uses Vertex Explainable AI and SHAP values to interpret model decisions. Builds stakeholder trust and satisfies transparency requirements.

Lesson 3 • AI Bias: Sources and Types

Identifies historical, representation, measurement, and aggregation bias in datasets and models. Establishes why bias detection must precede any deployment decision.

Lesson 4 • Fairness Metrics and Evaluation

Applies demographic parity, equalized odds, and individual fairness metrics to model outputs. Connects fairness measurement to business and legal accountability.

Lesson 1 • Retrieval-Augmented Generation

Grounds LLM responses in external knowledge bases using vector search and RAG pipelines. Reduces hallucinations and keeps answers factually current.

Lesson 2 • How Large Language Models Work

Explains transformer architecture, tokenization, and next-token prediction at a conceptual level. Provides the mental model needed to use and debug LLM behavior.

Lesson 3 • Prompt Engineering Techniques

Teaches zero-shot, few-shot, chain-of-thought, and role prompting strategies. Directly improves output quality without any model fine-tuning.

Lesson 4 • Gemini API Integration

Connects applications to Gemini models via REST and SDK calls with proper authentication. Covers multimodal inputs including text, image, and audio.

Lesson 1 • Measuring AI Business Impact

Defines KPIs, ROI frameworks, and A/B testing protocols for AI-driven features. Closes the loop between model performance and business outcomes.

Lesson 2 • Multi-Modal and Agentic AI Systems

Designs systems that combine text, image, audio, and tool-use agents for complex tasks. Represents the frontier of applied AI product development.

Lesson 3 • Cost Optimization for AI Workloads

Applies spot instances, model quantization, and caching to reduce AI infrastructure spend. Ensures AI projects remain financially viable at enterprise scale.

Lesson 4 • MLOps Maturity and Architecture

Assesses organizational MLOps maturity levels and designs target-state architectures. Bridges the gap between experimental models and reliable production systems.