Artificial Intelligence
·
Series A
·
2022
vision
AI vision stack for industrial safety compliance.
The problem
Industrial sites still rely on clipboards, spot checks and good intentions to enforce safety rules. Supervisors can’t watch every corner of a warehouse or construction site, so most incidents are only understood after someone gets hurt. Meanwhile, the cameras already installed on most sites record everything and are reviewed by no one unless there is an accident. Safety data existed, but it was locked inside hours of footage.
Why now
Industrial sites installed millions of security cameras over the past decade, but almost nobody watches the footage. Edge computing and efficient vision models now make it practical to analyse those feeds in real time without sending video to the cloud. Regulators and insurers are also placing more emphasis on proactive safety programs, while labour shortages make every injury more costly. Companies want measurable improvements, not just compliance paperwork. Vision arrived when the infrastructure was already in place, the technology was ready and the incentives to act were stronger than ever.
The founders
Rafael Costa worked as a safety engineer in heavy industry and investigated incidents that, in hindsight, had visible warning signs for weeks. Ingrid Holm led computer vision research at a robotics company and cared deeply about privacy. They designed Vision together around a principle that shaped everything: detect unsafe situations, not individuals. The system never stores faces, which made it acceptable to workers as well as managers.
Why we invested
Lina led our round because Vision’s privacy-first design made unions and safety officers equally enthusiastic, which is unusual for workplace monitoring. The product ran on existing cameras, required no new hardware and flagged hazards such as missing protective equipment, blocked exits or people too close to moving machinery in real time. Pilot customers saw measurable improvements within weeks. Lina was convinced the founders had built something that would become standard on industrial sites.
How we helped
We helped Vision navigate enterprise procurement at large industrial companies, introducing the founders to safety leaders and risk managers in our network. Lina worked with them on a pricing model based on site coverage rather than camera count, which simplified expansion. Our platform team connected them with insurers interested in using safety data to reward safer sites with lower premiums, creating a new channel for growth.
Where they are now
Vision now monitors hundreds of industrial sites, and customers report meaningful drops in recordable incidents within the first quarter of deployment. Several insurers now offer premium discounts to sites using the platform. The company recently added near-miss analysis that helps safety teams understand patterns before they lead to injuries. Rafael and Ingrid still review every serious alert with customers personally during their first month.

