NoSQL Database Schema Design for Scalable Applications

I'm building a [social media platform, IoT monitoring system, content delivery service, etc.] that needs to handle high volume data with flexible schema requirements. Help me design an optimal NoSQL database schema (preferably for MongoDB or DynamoDB) that will: 1) Support our core functionality including [describe 3-4 main features], 2) Scale efficiently to millions of users/devices, 3) Allow for fast queries on our most common access patterns, 4) Handle nested relationships effectively, 5) Support eventual consistency across distributed systems, and 6) Enable efficient analytics. For each collection/table, specify the document structure, key fields, indexes, potential for sharding, and denormalization strategies. Address concerns like preventing unbounded array growth, handling document size limits, and optimizing for read vs. write performance based on our use case. Also suggest a migration strategy if we were converting from a relational database.