Design NFT Minting Platform

Problem Statement: Creator NFT Minting Platform

An NFT minting platform lets creators launch collections and collectors mint tokens during staged drops — the OpenSea/Foundation/Manifold flow. The system mints ERC-721/1155 tokens on-chain, stores the art and metadata off-chain (IPFS/CDN), and runs an indexer that projects on-chain ownership into fast reads for explorer and wallet pages. It is mint-and-display focused, not a full secondary marketplace.

The defining architectural split — and what an interviewer wants stated early — is CP mint jobs vs AP metadata/CDN. The mint path is strongly consistent and money-adjacent: a mint must happen exactly once, with the right token going to the right wallet, after irrevocable on-chain confirmation. The metadata/media path is eventually consistent and read-heavy: images and JSON are served from IPFS + CDN where a little staleness is fine. Conflating these two is the classic mistake.

The scale to anchor: drops are spiky — a hyped collection mints out in minutes, so you size for the burst, not the average (thousands of concurrent minters slamming one collection). Confirmation depth gates when a mint is shown as final; the indexer's freshness (head − indexed block) is the read SLO.

The failure stories that define this domain are NFT-specific and worth naming: a metadata URI 404 after mint (token minted but its JSON/art is unreachable — a dead NFT), a duplicate tokenId from a race in the mint workers, a reveal leak before the countdown (collectors see the art early and snipe the rares), and royalty bypass on secondary sales. These, not a generic CRUD diagram, are what the role tests — alongside drop fairness (keeping bots from sweeping a drop).