The correct option is Multimodal Search because it lets you interpret and retrieve information from text in PDFs, visual content in images, and audio or video transcripts in a single unified search workflow.

With Multimodal Search, you can index heterogeneous assets and use shared embeddings to perform semantic retrieval across modalities. This capability returns relevant results whether the query or the answer lives in text, images, or recorded webinars, which directly matches the team’s goal of gathering insights across their internal asset library in one approach.

Document AI is designed for document understanding and extraction such as parsing PDFs and forms, yet it is not a unified cross modality search solution and does not provide end to end retrieval across images and recorded webinars.

Prompt Tuning is a parameter efficient way to adapt language models for specific tasks, but it does not provide indexing or retrieval and it is not a search capability across different media types.

Vertex AI Feature Store manages and serves machine learning features for training and online inference, and it does not offer content indexing or semantic search across documents, images, or videos.

Full AWS Practitioner Certification Question