Why this is here
Most of our client work is described without naming anyone, which makes it hard to judge how deep the engineering goes. RichIris is ours, it is open source, and anyone can go and look at it.
What it is
A self-hosted network video recorder. It records IP cameras continuously, indefinitely, and lets you scrub back through the footage.
- Recording. Continuous 24/7 capture with stream passthrough, so recording costs almost no CPU and no GPU at all.
- Playback. A zoomable 24-hour timeline with instant seek, variable speed in both directions, and hover-scrub thumbnail previews.
- Detection. Motion detection with per-camera sensitivity, plus on-device AI object detection distinguishing people, vehicles and animals, with multi-frame confirmation to suppress false positives.
- Facial recognition. Enrol known people from past detections; the timeline then distinguishes a known face from a stranger.
- Applications. Native Windows desktop and Android clients, not a web page in a wrapper.
- Operations. Runs as a Windows service with a health watchdog that detects and recovers from a silently dead listener, ships as a signed installer, and manages its own retention against age and disk limits.
Why it matters to a client
The same engineering shows up in commercial work: real-time data pipelines that cannot drop frames, AI inference running locally instead of shipping data to a third party, native applications on more than one platform, and software that survives being left alone for months.