AI Protex

We’re building a proactive computer vision tool for workplace safety – empowering industrial Health and Safety (H S) teams to identify risk and danger before it becomes a problem.

The Problem:
Keeping people safe at work is extremely hard. In the U.S. alone, there were over 5,000 deaths in the workplace in 2019, mainly in the heavy industry space. Serious workplace injuries cost U.S. business es circa $62 Billion annually. Until now, safety has been reactive, someone has to get injured, and in some cases fatally before something is done about it.

Problem Solved:
Protex AI’s platform uses privacy-preserving camera monitoring software to provide companies with an always-on guardian angel, protecting their workforce. The platform plugs into existing CCTV infrastructure and enables HS teams to flexibly translate their document-based safety rules into the real world. Protex AI takes these rules and autonomously audits the customer’s facility to identify areas of high risk and non-compliance. HS teams can access this data via a reporting engine and use it in weekly safety meetings, external safety audits, insurance discussions, and legal claims. The tool integrates seamlessly with existing safety workflows to augment data produced by any manual incident logging system in place.

Challenges:
The main challenges in building the platform centered around scalability and privacy. The Protex platform is backboned by an ever-evolving computer vision pipeline, it spans from the edge (native cloud) to the cloud, sitting within a hybrid architecture. This allows us to generalise our models and scale both within our customer sites and across new facilities. This pipeline consists of 5 stages: collection, labeling, training, evaluation, and deployment. Some of the most difficult challenges were presented in the data collection stage, namely, how do we find data that our model doesn’t understand? and how do we collect it while preserving customer privacy? To overcome this we developed a proprietary snapshotting infrastructure that collects data when there are discrepancies between our detector and tracker layers in our computer vision edge application. There are lots of other interesting parts of the system and we’re happy to share high-level overviews with the judges should this be required.

Results:
All of our customers have different problems when it comes to health and safety, but they all have the same goal, to keep their workforce safe. Using Protex AI, one of our customers was able to increase the number of incidents identified on their site by 60% leading to more productive data-driven discussions in their weekly safety meetings. Our customers are from the manufacturing, distribution, warehousing, and port industries and are adamant that the adoption of AI technology is the missing link in their health and safety workflow chain. “Protex AI solves a huge problem for our safety team. It’s simply impossible to keep an eye on everything that happens throughout the day on the terminal floor. Protex AI allows us to not only keep an eye out for risk but will also all ow us to understand any lapses in compliance” Mark Lawless, Operations Mana ger, Doyle Shipping Group.

Conclusion:
This category is titled “Best Application of AI in a Startup” – We believe that the ultimate application of AI is one that empowers people to do good while doing so in an ethical manner. Protex AI empowers health and safety teams to protect their workforce, allowing them to prevent accidents before they occur. Some of our customers have had multiple fatalities on their sit es in recent years and are using Protex to ensure that this never happens again. We put AI in the hands of the experts, allowing health and safety teams to leverage their domain knowledge by translating their document-based safety procedures into the real world.

Protex AI leverages the power of AI for good while also preserving organizational and individualistic privacy. We have strong privacy principles at Protex and are very proud of the privacy-preserving architecture that we’re building (edge computing and customer data control/ownership). Should we be shortlisted for these fantastic awards at the heart of the Irish AI community, we’d love to tell the judges more about it!