Google Cloud's Gen AI Offerings

Google has positioned itself as an AI-first company since 2016, when CEO Sundar Pichai declared the shift from "mobile-first" to "AI-first." This philosophy permeates every Google product — from Search and Gmail to Cloud and Android. For Google Cloud specifically, this means AI capabilities are embedded at every layer of the platform.

Google's AI advantages stem from three pillars: (1) world-class AI research from Google DeepMind and Google Research, which produced the Transformer architecture, BERT, and Gemini; (2) massive compute infrastructure with custom TPUs and hyperscale data centers; and (3) data at scale from billions of users across Search, YouTube, Gmail, and Maps.

Enterprise AI Platform

Google Cloud's enterprise AI strategy centers on Vertex AI as the unified platform. Rather than offering disconnected AI services, Google consolidates everything under Vertex AI: model training, deployment, fine-tuning, prompt engineering, RAG, agents, and monitoring. This "single pane of glass" approach simplifies enterprise adoption.

Training and serving large AI models requires specialized hardware. Google Cloud offers the most comprehensive AI infrastructure portfolio, including custom-designed TPUs, NVIDIA GPUs, and the Hypercomputer architecture that ties them together.

TPUs (Tensor Processing Units)

TPUs are Google's custom Application-Specific Integrated Circuits (ASICs) designed specifically for ML workloads. They excel at matrix multiplication operations that dominate neural network training and inference.

GPUs on Google Cloud

Google Cloud also provides NVIDIA GPUs for workloads that benefit from the CUDA ecosystem. The A3 GPU VMs (powered by NVIDIA H100 GPUs) are designed for large-scale AI training and inference. A2 GPU VMs (NVIDIA A100) serve general-purpose ML workloads. G2 GPU VMs (NVIDIA L4) are optimized for cost-efficient inference.

Google Cloud Hypercomputer

The Hypercomputer is Google's supercomputing architecture purpose-built for AI. It integrates:

• Optimized hardware — TPUs and GPUs with custom interconnects
• Open software — JAX, PyTorch, TensorFlow with XLA compiler optimization
• Flexible consumption — on-demand, reserved, and Dynamic Workload Scheduler
• Managed services — Vertex AI integrates directly with Hypercomputer resources

Google offers ready-to-use AI products powered by Gemini that require no development or infrastructure management. These are designed for end users and business professionals.

Gemini App, Gemini Advanced, and Gemini Enterprise

Gemini for Google Workspace

Gemini is deeply integrated into Google Workspace productivity tools, providing AI assistance directly in the applications people use every day:

Vertex AI Search

Vertex AI Search (formerly Enterprise Search) provides Google-quality search for enterprise data. It allows organizations to build search engines over their own documents, websites, and data stores with built-in AI understanding.

Key capabilities include:

• Multi-turn search — conversational follow-up queries that maintain context
• Extractive answers — pulls specific answers from documents, not just links
• Summarized answers — generates AI summaries from multiple sources
• Enterprise data connectors — indexes Cloud Storage, BigQuery, websites, Confluence, and more
• Access controls — respects existing data permissions and ACLs

Customer Engagement Suite (formerly CCAI)

The Customer Engagement Suite (formerly Contact Center AI) brings generative AI to customer service operations. It provides:

Vertex AI is Google Cloud's comprehensive AI development platform. It provides everything developers need to build, deploy, and manage generative AI applications. Think of it as the "IDE for AI" on Google Cloud.

Model Garden

Model Garden is a curated catalog of models available through Vertex AI. It provides a single interface to discover, test, and deploy models from multiple sources:

Vertex AI Studio

Vertex AI Studio is the interactive development environment within Vertex AI for prototyping and testing generative AI models. It provides:

• Prompt Design — visual interface for crafting and testing prompts
• Model Comparison — test the same prompt across multiple models side-by-side
• Parameter Tuning — adjust temperature, top-p, top-k in real-time
• System Instructions — set behavior guidelines for the model
• Grounding Configuration — connect to Google Search or custom data sources
• Code Export — generate Python/cURL code from your prompt configuration

Vertex AI Key Capabilities

Retrieval-Augmented Generation (RAG) is a technique that enhances generative AI by grounding model responses in external knowledge. Instead of relying solely on the model's training data, RAG retrieves relevant documents and includes them as context in the prompt.

Vertex AI provides managed RAG capabilities through multiple components:

# RAG with Vertex AI: simplified example
from vertexai.generative_models import GenerativeModel, Tool
from vertexai.preview import rag

# Create a RAG corpus from your documents
corpus = rag.create_corpus(display_name="company-docs")
rag.import_files(corpus.name, paths=["gs://my-bucket/docs/"])

# Query with grounding
model = GenerativeModel("gemini-1.5-pro")
rag_tool = Tool.from_retrieval(
    retrieval=rag.Retrieval(
        source=rag.VertexRagStore(rag_corpora=[corpus.name])
    )
)
response = model.generate_content(
    "What is our company's refund policy?",
    tools=[rag_tool]
)

Vertex AI Agent Builder is a platform for creating AI agents — autonomous systems that can reason, plan, use tools, and take actions to accomplish goals. Agents go beyond simple Q&A by interacting with external systems and making multi-step decisions.

Agent Builder supports multiple agent types:

Agent Tooling and Extensions

Agents are only as powerful as the tools they can use. Vertex AI provides a rich tooling ecosystem:

# Function calling example with Gemini
from vertexai.generative_models import GenerativeModel, FunctionDeclaration, Tool

# Define a function the model can call
get_weather = FunctionDeclaration(
    name="get_weather",
    description="Get current weather for a city",
    parameters={
        "type": "object",
        "properties": {
            "city": {"type": "string", "description": "City name"}
        },
        "required": ["city"]
    }
)

weather_tool = Tool(function_declarations=[get_weather])
model = GenerativeModel("gemini-1.5-pro", tools=[weather_tool])
response = model.generate_content("What's the weather in Tokyo?")
# Model returns: function_call { name: "get_weather" args: { city: "Tokyo" } }

A major theme of the exam is matching scenarios to the correct Google Cloud offering. Use this decision framework: