Vertex AI
Unified enterprise AI platform for training, tuning, and deploying generative AI and custom ML models.
Unified enterprise AI platform for training, tuning, and deploying generative AI and custom ML models.
What is Vertex AI?
Google Cloud Vertex AI is a unified machine learning platform that brings all of Google's AI tools, foundation models (Gemini 1.5 Pro/Flash, Imagen, Codey), and MLOps infrastructure into a single integrated console and API. From prompt engineering in Vertex AI Studio to custom distributed neural network training on NVIDIA H100 GPUs and Cloud TPUs, Vertex AI provides enterprise governance, vector embeddings search, and real-time model serving.
Decision Guide: When to Choose This Service
Use Vertex AI for enterprise GenAI applications (Gemini models, Model Garden, vector search embeddings), training custom TensorFlow/PyTorch/scikit-learn models on GPUs/TPUs, feature stores, and production model serving with autoscaling endpoints.
Do NOT use Vertex AI if you only need simple SQL-based tabular predictions on BigQuery data (use BigQuery ML instead without leaving SQL).
Core Architectural Pillars
Interactive playground for prompting and fine-tuning foundation models (Gemini 1.5, Anthropic Claude, Llama 3, Mistral).
Model Registry versions trained model artifacts; Endpoints provide autoscaling HTTPS REST/gRPC prediction endpoints.
Petabyte-scale vector database for semantic search and Retrieval Augmented Generation (RAG).
Centralized repository to serve, share, and prevent training-serving skew in ML feature values.
Anchor GenAI responses in real-time factual data from Google Search or private BigQuery/GCS repositories.