Generative AI Foundations for Cloud

Complete overview of Generative AI (GenAI) services across major cloud platforms: AWS, Azure, Google Cloud, Alibaba Cloud, NVIDIA, and Salesforce.

Table of Contents

1. Introduction
2. Overview — GenAI for Cloud
3. AWS — Generative AI
4. Azure — Generative AI
5. Google Cloud — Generative AI
6. Other Cloud Platforms
7. Comparative Summary

1. Introduction

This course explores the GenAI capabilities of major cloud providers:

• AWS (Amazon Web Services)
• Microsoft Azure
• Google Cloud Platform (GCP)
• Alibaba Cloud
• NVIDIA
• Salesforce

2. Overview — GenAI for Cloud

2.1 Defining Generative AI

Generative AI: technology that creates new content from existing data.

GenAI for cloud is technology that creates new platform-specific content from existing data — where the data may be platform-restricted or open.

2.2 Main GenAI Service Categories

3. AWS — Generative AI

3.1 AWS GenAI Architecture (3 Levels)

graph TD
    subgraph L3["Level 3 — Applications"]
        Q[Amazon Q\nGenAI Assistant]
        QCW[Q in CodeWhisperer\nCode assistance]
    end

    subgraph L2["Level 2 — Tools"]
        BED[Amazon Bedrock\nUnified access to Foundation Models]
    end

    subgraph L1["Level 1 — Infrastructure"]
        SM[Amazon SageMaker\nBuild / Train / Deploy ML]
        TR[Trainium2\nAWS training chip]
        SM --- TR
    end

    L1 --> L2 --> L3

3.2 Amazon SageMaker

Fully managed service for building, training, and deploying machine learning models at scale.

SageMaker JumpStart:

• Collection of pre-built foundation models
• Usable as-is or fine-tunable with your data
• Hundreds of ML algorithms covering: classification, sentiment analysis, fraud prediction

SageMaker Canvas — No-code Workflow:

flowchart LR
    S1["1. Data selection\n(natural language UI)"]
    S2["2. Model building\n(AutoML)"]
    S3["3. Accuracy evaluation\n(test analysis)"]
    S4["4. Deployment\n(SageMaker endpoint)"]
    S1 --> S2 --> S3 --> S4

Security: Your organization’s data is not used to pre-train models. Custom models are separate instances. All data is encrypted and never leaves your VPC.

3.3 Amazon Bedrock

Exposes foundation models as a fully managed service via a single API.

graph TD
    subgraph API["Single API Access"]
        M1[Anthropic Claude]
        M2[Meta Llama 2]
        M3[AI21 Labs Jurassic]
        M4[Cohere Command & Embed]
        M5[Stability AI Stable Diffusion]
        M6[Amazon Titan]
    end

    USER[Developer / Application] -->|API Call| BED[Amazon Bedrock]
    BED --> API
    BED --> AG[Bedrock Agents\nMulti-step tasks]
    BED --> GR[Guardrails\nAI & security policies]

Guardrails Bedrock:

3.4 Amazon Q

Amazon Q is AWS’s GenAI assistant, configured according to your organization’s roles and permissions.

Amazon Q Integrations:

• Q Business — 40+ enterprise connectors (S3, Salesforce, Google Drive, Microsoft 365, ServiceNow, Slack)
• Q for AWS — AWS cloud expertise
• Q in QuickSight — Business Intelligence
• Q in Connect — Customer service
• Q in CodeWhisperer — Code assistance

3.5 Amazon CodeWhisperer

AWS AI code assistant trained on open source and Amazon code.

4. Azure — Generative AI

4.1 Azure GenAI Architecture (3-Level Building)

graph TD
    subgraph SR["Showroom — Accessibility"]
        ACS[Azure Cognitive Services]
        ABS[Azure Bot Service]
    end

    subgraph WS["Workshop — Customization"]
        AML[Azure Machine Learning]
        AIST[Azure AI Studio]
    end

    subgraph ER["Engine Room — Powerful Models"]
        AOAI[Azure OpenAI Service\nChatGPT · DALL-E · Codex]
    end

    ER --> WS --> SR

4.2 Azure OpenAI Service

Gateway to OpenAI’s foundation models, including ChatGPT versions.

4.3 Azure AI Studio

Enables application developers (without data science expertise) to create GenAI applications and custom Copilot experiences.

Available models:

• Mistral (multilingual mastery)
• Azure OpenAI (code, analysis, creative text)
• Hugging Face (community LLM repository — the “GitHub of LLMs”)
• Meta (conversational interfaces)
• Deci (optimization & efficiency)

4.4 Additional Azure AI Services

4.5 Azure AI Content Safety & Responsible AI

Azure AI Content Safety (in EN, DE, ES, FR, PT, IT, ZH):

• Analyze Text API — Sexual · Violence · Hate · Self-harm
• Analyze Image API — Sexual · Violence · Hate · Self-harm
• Jailbreak Risk Detection API
• Protected Material API

Azure Responsible AI — 6 Principles:

5. Google Cloud — Generative AI

5.1 Google Cloud GenAI Architecture

graph TD
    subgraph APP["Applications"]
        DA[Duet AI\nMulti-service AI assistant]
    end

    subgraph TK["Toolkit"]
        VAIS[Vertex AI Studio\nSimplified no-code interface]
        API2[Vertex AI APIs]
    end

    subgraph FND["Foundation"]
        VAI[Vertex AI\nGlobal ML platform]
        MG[Model Garden\n130+ models]
        GEM[Gemini\nTop-of-line multimodal model]
        PALM[PaLM 2\nComplex text LLM]
        VAI --> MG
        MG --> GEM
        MG --> PALM
    end

    FND --> TK --> APP

5.2 Google Cloud Gemini

Multimodal foundation model supporting native multimodal prompting (text, code, image, video).

Gemini vs Imagen vs Codey

5.3 Google Cloud Model Garden

Central library of pre-trained models with 130+ models available.

4 access routes:

• Direct API
• Vertex AI Studio (no-code)
• Jupyter Notebooks
• Vertex AI Pipeline (serverless)

5.4 Vertex AI Studio

Simplified console interface for GenAI, accessible even without ML expertise.

Supported prompt types:

5.5 Vertex AI Search and Conversation

Vertex AI Search — custom semantic search applications:

• Semantic understanding (NLP + Google Search expertise)
• Source flexibility (structured data, documents, support content)
• Customization (rankings, synonyms, ML schemas)

Vertex AI Conversation — GenAI-powered chat and voice bots:

• Personalized product recommendations
• Complex support & helpdesk
• Interactive training with adaptive content

5.6 Google Cloud Duet AI

Multi-faceted Google Cloud AI assistant:

6. Other Cloud Platforms

Alibaba Cloud

NVIDIA

Salesforce

7. Comparative Summary

Service Comparison by Platform

Responsible AI — Approach Comparison

The GenAI cloud landscape evolves extremely rapidly. Consult each provider’s official documentation for the latest updates.

Search Terms

generative · ai · foundations · cloud · genai · artificial · intelligence · azure · google · amazon · architecture · service · aws · comparison · gemini · responsible · studio · vertex