How Protex AI Builds Fast and Saves Lives

Ireland-based Protex AI uses computer vision to help enterprise workplace health and safety teams make proactive safety decisions contributing to a safer workplace. They build on the Hasura platform while protecting data privacy with granular, role-based authorization.

The Problem: Protex AI’s Challenge

Workplace health and safety is a matter of life and death; in the U.S., a worker died every 101 minutes in 2021.

Ciarán O’Mara, the Co-Founder and CTO of Protex AI, heard of a death on a production line and wondered whether AI and machine learning could help prevent such accidents. O’Mara and long-time friend Dan Hobbs started interviewing health and safety practitioners, speaking with 200 people before writing a single line of code.

“Companies either take health and safety seriously or very seriously,” says O’Mara. “There's no one that does not take health and safety seriously.”

Safety challenges cut across industries, from manufacturing to logistics, meat processing, energy, oil, gas, and more. Despite ongoing investments in best practices, specialized equipment, and training, accidents and fatalities continue.

Closed-circuit television (CCTV) cameras monitor production line activity from all angles. What if those video feeds could be analyzed to find predictors or indications that best practices were slipping and risk was increasing?

But running machine learning on CCTV video in a way that does not violate privacy laws presents yet another challenge. Any solution must balance safety with privacy, satisfying both health and safety professionals and the IT leaders who scrutinize technology solutions.

Protex AI’s solution would need to complement existing technology stacks mainly made of Excel, email, Microsoft Teams, and reporting software. “We would need to sit on top of that stack and augment the data pool that health and safety people have to make decisions,” said O’Mara.

The company’s goal? To deliver value to health and safety teams week after week, providing insights and education that materially improve the health and safety of employees.

The Solution: Why Protex AI Chose Hasura

O’Mara and another developer set about building a proof of concept (POC) with an Express.js backend, which required writing microservices from the ground up.

“There were two of us building the proof of concept. We knew it wasn't going to scale, or if it was going to scale, it was going to be painful,” said O’Mara. “We were putting a lot of time into architecting APIs when we wanted to put the time into architecting the database.”

An intern showed them GraphQL, which led them to Hasura, with its instant API capabilities. “It became obvious straight away that Hasura was going to help us build fast,” said O’Mara. “With Hasura, the API piece becomes an abstracted afterthought. Hasura writes the resolvers so we don’t have to.”

When developers had questions, the Hasura Community offered immediate answers. “Having the ability to chat with Hasura engineers was massive,” said O’Mara.

As the Protex AI team continued to build with Hasura, they found even more reasons to make it their development platform of choice.

Diverse Development Teams Meet in Hasura

Protex AI has several development teams: a product team, an infrastructure and security team, a computer vision and systems team, a data and ML ops team, and a privacy and security team. Hasura is a unified surface where the teams meet.

“Hasura is the same language to the integrators on the computer vision team and the product team,” said O’Mara. “When we think about our database or API system, it's Hasura, even though we have a full database stack behind it.”

Accelerating Onboarding

Hasura cuts the time for onboarding developers. “When we onboard a new hire, we give them a couple of projects centered around Hasura,” said O’Mara. “They're able to pull data and play with it, show it on the front end, and see that it works.”

The Hasura API Explorer makes it easy for Protex AI developers to build events and visualize what they're doing. Time to productivity for new developers is impressive: “It takes us less than two weeks to get new engineers up and running using the API system,” says O’Mara.

Satisfying Security Requirements

Protex AI has to satisfy its customer’s IT security leaders, who are concerned with data security, as well as the privacy of the CCTV camera feeds that Protex AI ingests and analyzes.

Hasura’s granular authorization enables Protex AI to lock down access to data. Protex AI deployed Hasura Enterprise on-premise to oversee sensitive customer data flowing through their database. Hasura’s security features have helped Protex AI in passing ISO27001 audits as well as penetration tests.

The Results: The benefits of using Hasura

Hasura’s instant API capabilities do the lion’s share of the work in creating APIs. There’s no need to hire API engineers because Hasura solves the problem for them. Without Hasura, O’Mara estimates that they would need two additional experienced API architects to manage the extra workload.

“Hasura has allowed us to stop thinking about data access and start thinking about data processing,” says O’Mara. “It's data abstracted for us. We don't have to worry about data connectivity.”

Recent penetration tests showed that authorization rules written in Hasura blocked unauthorized access to the data in Protex AI’s database. Pen testers could see front-end components but could not get to the data. Importantly, pen testers attempted to exploit the OWASP Top 10 security vulnerabilities, and Protex AI successfully passed those tests.

Building fast helps Protex AI close business. During a POC, a new customer may request several features. Delivering some features immediately strengthens the business case and often wins the deal.  “Building on Hasura has enabled us to ship features that require no database changes and build internal tools in less than a week in some cases,” says O’Mara.

Protex AI picks up on signals such as lapsing safety behaviors and procedures that aren’t being followed. “That is the way that health and safety should be practiced,” says O’Mara. “It's picking up on things before they lead to accidents as opposed to waiting for accidents to happen.”

Hasura frees up developers so that Protex AI can focus on its business in much the same way as they use computer vision tools instead of writing them.

The company can then focus on its core mission: Making the world a safer place.

Ready to hear more interesting and insightful real-life Hasura use cases? Register today for HasuraCon 2023, happening June 20-23, both virtual and free!

31 May, 2023
Subscribe to stay up-to-date on all things Hasura. One newsletter, once a month.
Accelerate development and data access with radically reduced complexity.