Principle i / The proof / Sazmining, 2024–present

Build what works in the real world.

Dispatch across five facilitiesCommands leave a Redis queue, pass through a scheduler, and arrive at five facilities, throttled at each one.01 · QUEUERedis, in order02 · SCHEDULERsends everywhere03 · THROTTLE, EACH FACILITYDC-01DC-02DC-03DC-04DC-05

The problem

Customers own hardware hosted across multiple data centers. Each machine needs commands, and a burst of them at one site causes power fluctuations.

The decision

Treat power as a real constraint. A Redis queue plus a scheduler sends commands to machines everywhere, throttled for each facility so the load stays even.

The conditions

The company grew 350% in 2025, from 2 to 5 data centers. For a time I was the only engineer, while also splitting the monolith into Admin, Customer and commerce apps plus a QuickBooks service.

Architected the migration of a monolith into three purpose-built applications (Admin, Customer, e-commerce), including a QuickBooks microservice and an automated billing pipeline on Supabase Edge Functions, restructuring how financial operations run across subscription and e-commerce systems.

Sazmining is a hardware hosting and managed infrastructure service.

Software that touches the physical world has to respect it.

ii. Own the whole thing →