Constructing a scalable doc administration system: Classes from separating metadata and content material

{
  "unique_document_id": "aGVsbG93b3JsZA==",
  "member_id": "123456",
  "file_name": "employment_contract.pdf",
  "document_category_id": 101,
  "document_subcategory_id": 10110,
  "document_extension": ".pdf",
  "document_size_in_bytes": 245678,
  "date_added": "2025-09-20T12:11:01Z",
  "date_updated": "2025-09-21T15:22:00Z",
  "created_by_user_id": "u-01",
  "updated_by_user_id": "u-02",
  "notes": "Signed by each events"
}

For question patterns, I leveraged secondary indexes aggressively. Whereas the first desk makes use of the distinctive doc ID as its key, a secondary index organized by member ID and doc class permits environment friendly queries like “retrieve all paperwork of a sure class for a given member” with out costly desk scans.

The schema-on-read mannequin of NoSQL proved invaluable for evolution. After we wanted so as to add a brand new non-compulsory metadata discipline, there was no dangerous ALTER TABLE assertion or downtime. New paperwork merely began together with the attribute, whereas present paperwork continued working with out it. This agility allowed us to reply to new necessities in hours as a substitute of weeks.

Constructing in catastrophe restoration and knowledge resiliency

A complete catastrophe restoration technique was important for enterprise continuity. I included resiliency at each the metadata and content material layers.