AWS Certified Cloud Practitioner

Explanation

Amazon Bedrock is the BEST fit for this scenario because:

1. Single Unified API: Amazon Bedrock provides access to multiple foundation models (FMs) from leading AI companies through a single API, which aligns perfectly with the requirement for a "single unified API."

2. Multiple Capabilities: Bedrock supports text generation (for product descriptions), image generation (for marketing creatives), and conversational AI (for chatbot features) - all three use cases mentioned in the question.

3. No Infrastructure Management: Bedrock is a fully managed service, so the company doesn't need to manage GPUs or scale infrastructure, which matches their requirement to "avoid managing GPUs or scaling infrastructure."

4. Serverless Experience: Bedrock offers a serverless experience where AWS handles all the underlying infrastructure, scaling, and maintenance.

Why the other options are not the best fit:

A. Build separate microservices using Amazon Comprehend, Rekognition, and Polly:

• These are individual AI services for specific tasks (Comprehend for NLP, Rekognition for computer vision, Polly for text-to-speech)

• They don't provide a "single unified API" - you'd need to integrate multiple services separately

• They don't cover all the required capabilities (Polly is for speech synthesis, not chatbot functionality)

B. Use Amazon SageMaker to train custom models for each use case:

• Requires significant ML expertise and infrastructure management

• Involves training, deploying, and scaling models - which contradicts the requirement to avoid managing infrastructure

• Not a "single unified API" solution

D. Host open-source models in EKS with autoscaling:

• Requires managing Kubernetes clusters and infrastructure

• Involves GPU management and scaling challenges

• Not a unified API solution and requires significant operational overhead

Amazon Bedrock's serverless nature, unified API, and access to multiple foundation models make it the ideal choice for companies wanting to leverage generative AI capabilities without managing infrastructure.