Question 1: Multimodal RAG Retrieval Question In a Retrieval-Augmented Generation (RAG) system, how would you efficiently handle and retrieve information from multiple types of datasets such as text, images, audio, video, and tables? For example: Text documents may contain paragraphs, structured tables, and columns Images may contain charts or diagrams Audio and video may contain spoken content How would you design the data processing and retrieval pipeline so that the system can retrieve the most relevant information efficiently when a user asks a question? Specifically explain: How each modality (text, tables, images, audio, video) would be processed How the data would be converted into embeddings How it would be stored in a vector database How the system would perform efficient retrieval across different data types Question 2: Resume Information Extraction Question You are given multiple resume templates, and each template contains different formats of date representations. Examples of date formats include: Jan 2022 – Mar 2023 2021 - Present 03/2020 – 07/2022 March 2019 to June 2021 Your task is to build a system that automatically extracts structured information from resumes, specifically: Project Name Project Duration Total Years of Experience Challenges: Resumes follow different templates and layouts Date formats are not consistent Information may appear in different sections