AI Model Wiki

    Skip to main content
    20 min
    Lesson 1 of 4

    Vertex AI Architecture

    From the course Google Vertex AI Platform

    Google Vertex AI is a unified ML platform combining managed infrastructure with access to Google's foundation models.

    Platform Components

    Model Garden

    • Gemini models: Google's flagship LLM family
    • Third-party models: Anthropic Claude, Meta Llama
    • Open models: Mistral, Falcon, other open-source
    • Specialty models: Code generation, translation, embeddings

    Vertex AI Studio

    • Web-based playground for model testing
    • Prompt engineering and management
    • Evaluation tools and benchmarking
    • Fine-tuning interface

    AutoML

    • Train custom models without code
    • Supports: tabular, image, text, video data
    • Automated feature engineering
    • Hyperparameter optimization

    Pipelines

    • Orchestrate end-to-end ML workflows
    • Kubeflow Pipelines or TFX backends
    • Reproducible, versioned experiments
    • Artifact tracking and lineage

    Gemini via Vertex AI

    import vertexai
    from vertexai.generative_models import GenerativeModel
    
    vertexai.init(project="my-project", location="us-central1")
    
    model = GenerativeModel("gemini-2.5-pro")
    
    response = model.generate_content(
        "Explain the difference between RAG and fine-tuning for LLMs",
        generation_config={
            "max_output_tokens": 1024,
            "temperature": 0.7,
            "top_p": 0.9
        }
    )
    
    print(response.text)
    

    Multimodal Capabilities

    from vertexai.generative_models import Part, GenerativeModel
    
    model = GenerativeModel("gemini-2.5-pro")
    
    # Process image + text together
    response = model.generate_content([
        Part.from_uri("gs://my-bucket/product.jpg", mime_type="image/jpeg"),
        "Describe this product and suggest marketing copy for it"
    ])
    
    print(response.text)
    

    Key Takeaways

    • •Vertex AI unifies model access, training, deployment, and MLOps in one platform
    • •Model Garden provides access to Gemini, Claude, Llama, and open-source models
    • •AutoML enables custom model training without machine learning expertise
    • •Gemini 2.5 Pro handles text, image, audio, and video inputs natively
    0 of 4 lessons marked complete

    Progress is kept in this browser only.