Connecting camera management to tenant deployment

ISP Hyperdrive is a team-built camera platform. Through Centreville Tech, I built the entire tenant onboarding system and contributed camera-management workflows, along with scheduling and upload-recovery changes in its background workers. As of October 2026, I estimate the platform footprint at 3 tenants and approximately 25 cameras.

My control-plane work connects onboarding and camera configuration with the deployment configuration a tenant needs. I also worked on making deployment rendering come from a canonical source, reducing the number of places that need to agree when a configuration changes.

Scheduling example · hourly work behind a monthly backlog
BeforeLong frame-download batches for weekly or monthly videos could delay hourly timelapses.
DecisionTrack each camera and window independently; reconsider the queue after each job and choose shorter due windows first.
RecoveryDefer lengthy downloads, retain cached frames, and resume later without publishing a partial video.

A regression test puts monthly work ten days overdue and still expects the due hourly job first.

Choose freshness without lowering video quality

I revised the scheduler from ranking relative lateness to prioritizing shorter due windows. This favors fresh hourly output over clearing the oldest long-window backlog. Each camera and window keeps its own due time, so finishing one job does not advance another camera's schedule.

Frame synchronization can yield between downloads and resume from cache on a later attempt. Deferred work is not encoded or uploaded as an incomplete video. Policy tests cover backlog priority, independent camera schedules, retry timing, and unchanged video-quality settings.

The tradeoff is explicit: shorter windows get priority, while long-window completion still depends on available capacity. My work also includes upload-recovery handling; the broader platform was built with other contributors.