Perception Engineer (Computer Vision)

Rimward builds software that turns existing camera infrastructure into low-noise, operator-ready intrusion alerts for remote outdoor industrial sites.

The company is led by F-G Fernandez, a repeat founder with a strong computer vision background, who previously built and deployed vision systems in the field with Pyronear. We are now applying that same field-first mindset to physical security: making detection systems that stay useful under real outdoor conditions, not just in curated datasets or demo environments.

We are looking for a perception engineer to push detection quality on real-world outdoor video: intrusion detection, low false positives, difficult lighting, long-range scenes, PTZ cameras, and edge-constrained deployments.

You will work on the vision side of the stack, but not in isolation. We want someone who understands that real product performance is not just model metrics; it is behavior in noisy environments, under deployment constraints, with customer trust on the line. The hard part is not just finding something in a frame. It is deciding what is worth escalating on a real site, under real conditions, with a false-positive budget that operators can live with.

Within your first 30 days, you should be able to understand the current detection pipeline, failure modes, and evaluation setup, and start improving one concrete weak spot in production-like data. Within 60 days, you should be running disciplined experiments tied to real deployment constraints: lighting shifts, scene-specific noise, camera motion, or edge-performance tradeoffs. Within 90 days, you should own meaningful improvements in detection quality, evaluation rigor, and the feedback loop between model behavior and what operators actually trust in the field.

This role is for someone who can combine strong computer vision fundamentals with practical engineering and sharp experimental discipline. You should enjoy messy, high-consequence problems where the right answer is not just a better benchmark score, but a system that behaves reliably on difficult outdoor sites.

Interview Process

1. Intro call (30 min)
2. Technical deep dive (60 min)
3. AI-assisted analysis exercise (2–4 hours max)
4. Applied vision interview (60 min)
5. Founder final (45 min)