AutoML (Vertex AI)
Train high-quality custom machine learning models with zero data science coding required.
Train high-quality custom machine learning models with zero data science coding required.
What is AutoML (Vertex AI)?
AutoML on Vertex AI enables developers with minimal machine learning expertise to train high-accuracy custom models. Leveraging Google's Neural Architecture Search (NAS) and transfer learning, AutoML explores thousands of neural architectures in parallel to discover the optimal model for your tabular, image, video, or text datasets.
Decision Guide: When to Choose This Service
Use AutoML when you have labeled domain-specific data (custom product images, specialized text classification, tabular churn prediction) and want state-of-the-art machine learning models trained automatically without writing model training code.
Do NOT use AutoML if off-the-shelf pre-trained APIs (Cloud Vision, Natural Language) already meet your accuracy needs, or if you need to hand-tune deep neural network architectures in PyTorch.
Core Architectural Pillars
Image (classification, object detection), Tabular (regression, classification), Text (entity extraction, sentiment), and Video (action recognition).
Google algorithms that design custom neural networks tailored specifically to your data.
Generates feature attribution charts showing which inputs most heavily influenced the model's predictions.
Export trained AutoML vision models to TensorFlow Lite or ONNX to run offline on mobile and edge devices.