What you'll learn

Use text embeddings to capture the meaning of sentences and paragraphs
Apply text embeddings for tasks like text clustering, classification, and outlier detection
Use Google Cloud’s Vertex AI to build a question answering system

Course recap

PRO

This course teaches how to use Google Cloud's Vertex AI platform for building generative AI applications. It covers text embeddings, text generation, and how to leverage Google's cloud infrastructure for deploying and managing AI models. Students learn to work with embedding APIs, understand embedding properties, and apply them to practical tasks like classification, clustering, and semantic search.

Concept map

Key concepts

EmbeddingsVector representations capturing semantic meaning, computed via transformer neural networks with contrastive learning
Text embeddingsSubword token-level embeddings averaged to sentence-level representations, enabling semantic similarity computation
Contrastive learningTraining technique that moves similar sentence embeddings closer and dissimilar ones apart
Multimodal embeddingsEmbedding both text and images into the same vector space (e.g., CLIP)
Semantic searchUsing embedding similarity to find relevant documents or answers

About this course

The Vertex AI Text-Embeddings API enhances the process of generating text embeddings. These text embeddings, which are numerical representations of text, play a pivotal role in many tasks involving the identification of similar items, like Google searches, online shopping recommendations, and personalized music suggestions.

During this course, you’ll use text embeddings for tasks like classification, outlier detection, text clustering and semantic search. You’ll combine semantic search with the text generation capabilities of an LLM to build a question-answering systems using Google Cloud’s Vertex AI.

You’ll also explore:

The properties of word and sentence embeddings
How embeddings can be used to measure the semantic similarity between two pieces of text
How to apply text embeddings for tasks such as classification, clustering, and outlier detection
Modify the text generation behavior of an LLM by adjusting the parameters temperature, top-k, and top-p
How to apply the open source ScaNN (Scalable Nearest Neighbors) library for efficient semantic search
How to build a Q&A system by combining semantic search with an LLM

Upon successful completion of this course, you will grasp the underlying concepts of using text embeddings, and will also gain proficiency in generating embeddings and integrating them into common LLM applications.