Master Vertex AI: Leveraging LLMs with Text-Embeddings API

Set up your development environment with Python, VS Code, and a Google Cloud project to access Vertex AI text embeddings, understand API keys, costs, and the free 90 day tier.

Set up Google Cloud by creating a project, enabling Vertex AI, BigQuery, and IAM services, and generating a service account with a JSON key for authentication.

Set up credentials and a JSON file, install the Google Cloud AI platform, and initialize Vertex AI with your project and region. Test a text embedding to verify the setup.

Discover how Vertex AI generates text embeddings and enables similarity search, classification, clustering, and outlier detection. Learn about its embeddings API and open platform for scalable AI.

Learn how embeddings convert text and images into a continuous dense vector space where semantically similar items, like cat and kitty, stay close, enabling efficient processing and better generalization.

Explore how embeddings empower GenAI and LLMs to understand context, generate coherent text, and enable semantic search, text classification, recommendations, and outlier detection.

Discover Vertex AI embeddings API: text embeddings turn text into 768-dimensional vectors for semantic search, recommendations, and RAG, while multimodal embeddings unify text, images, and video for cross-modal search.

Explore how vertex ai task types optimize embeddings for specific use cases, improving document retrieval, question answering, and fact verification with models like text embedding Geico 003.

Explore how text, image, and video inputs map to a unified multimodal embedding space using Vertex AI. Apply cross-modal search and multimodal embeddings to enable unified understanding across data types.

Explore how to generate text embeddings with a pretrained model in Vertex AI, inspect the embedding vectors, and verify a 768-dimensional feature space while experimenting with different texts.

Explore embeddings with Vertex AI, install scikit-learn, and compute cosine similarity to compare sentence semantics in a vector space.

Visualize text embeddings by converting sentences to 768-dimensional vectors, reduce to 2D with PCA, and plot interactive 2D visuals using numpy, matplotlib, and mpl cursors.

Build practical skills with Vertex AI and the text embeddings API by generating embeddings, applying cosine similarity, and visualizing embeddings in the vector space for text and multimodal data.

Pablo invites you to check in, asks for course reviews to help others see the value, and encourages posting questions on the discussion board for help and community answers.

Explore how to generate and summarize text with Vertex AI using the text generation model powered by the text bison pre-trained model, including prompts, predictions, and output metadata.

Master text generation and classification with prompts in a large language model, tuning outputs via max tokens, temperature, top K and top P, and apply to remote team decisions.

Extract information from text into a table or json by using prompts and the predict function to capture team members and their roles.

Learn to control model behavior by adjusting temperature, top K, and top P; compare deterministic outputs at zero to creative outputs near one using hands-on prompts and simple examples.

Experiment with top-k and top-p sampling to balance diversity and coherence in text generation, adjusting temperature to observe how prompts produce different outputs.

Learn to summarize and extract insights from transcripts with Vertex AI's bison model, performing sentiment analysis and extracting next steps in bullet points from JSON transcripts.

Explore real-world data retrieval with BigQuery to fetch Stack Overflow questions and answers for Python, Java, and Dart; generate text embeddings and visualize 2D clusters.

Builds a retrieval augmented generation rag using StackOverflow data, local embeddings, and semantic search to answer questions with an llm, showcasing doc retrieval and prompt design.

Compare approximate nearest neighbor search with hnsw versus cosine similarity using the face library, showing faster latency (about 4.7 ms vs 8.7 ms) on large data sets.

Explore Vertex AI and the text embeddings API, covering embeddings, text and multimodal embeddings, embedding task types, cosine similarity, and visualization, plus using BigQuery to create and compare embeddings.