← All selected work

Performance and cost

Health-data storage redesigned around access patterns

Moved the right sensor payloads out of the primary database while preserving the product workflows that depended on them.

Primary evidence75%

less database storage

Situation

What the system needed to change

A health-tech platform serving more than 150,000 users was retaining large sensor payloads in its primary database, increasing disk use and operational cost.

Constraints

The boundaries mattered.

  • 01

    Existing product behavior could not break during migration

  • 02

    Large asynchronous payloads still needed reliable processing

  • 03

    The solution had to remain understandable to the product team

Decisions

The important engineering choices

  • Separated frequently queried relational state from large object payloads
  • Moved suitable data to S3-backed storage
  • Kept the database as the source of truth for product-facing metadata

Outcome

What the work left behind

  • Reduced database disk usage by 75 percent
  • Preserved product access to historical sensor data
  • Created a more durable boundary between transactional and object storage

Start with the problem

Have a system with similar constraints?

Send the workflow and the boundary that makes it difficult. The first conversation is about fit, not a generic sales deck.