The Valais-based startup Neuralia, headquartered in Sion, and the SME Services Techniques Alpins SA (STA), which specializes in industrial equipment maintenance in Sembranch, have collaborated on an innovative project supported by The Ark Foundation. The goal: to develop an artificial intelligence solution capable of assisting technicians in diagnosing breakdowns by leveraging service history data and technical manuals. With an average rating of 3.4/5 from operators during testing, this prototype demonstrates AI’s potential to preserve technical expertise and accelerate the skill development of new employees.
Like many companies active in technical maintenance and repair, STA faces a critical skills shortage. “Senior technicians possess valuable expertise, but this expertise is gradually disappearing as they leave,” explains Lionel May, director of STA. “Conversely, junior technicians struggle to quickly develop their skills and effectively diagnose breakdowns early in their careers.”
Faced with this reality, STA sought a solution to leverage data from breakdowns to assist its technicians in their diagnostics. “We needed to find a way to capitalize on the experience we’ve accumulated over the years,” says Lionel May. “The idea was to transform this knowledge into a practical tool accessible to all our technicians.”
It was in this context that the project with Neuralia began, with the goal of combining service history, technical manuals, and artificial intelligence to create a diagnostic assistant.
Three Main Objectives
The project had three main objectives, according to Gary Folli, founder of Neuralia. “First, to troubleshoot faster and more effectively by guiding technicians to the right solution at the right time. Second, to preserve expertise before it disappears by capturing the know-how of senior technicians. Third, to accelerate the skill development of junior technicians, enabling them to become more independent more quickly.”
For Lionel May, the stakes were particularly high in the current context. “The shortage of skilled labor is forcing us to rethink our work methods. We can no longer afford to lose years of expertise every time someone retires.”
The project began with a needs assessment, followed by a proof of concept (PoC) carried out with the support of The Ark. “We utilized the fault databases as well as the machine manuals, which we combined with AI,” explains Gary Folli.
Extensive Testing with Operators
Once the prototype was developed, it was extensively tested by STA operators. “Technicians could leave notes and comments on the responses generated by the AI. “This feedback allowed us to gauge satisfaction and identify areas for improvement,” explains Gary Folli.
In total, 53 failure cases were tested, with an average rating of 3.4 out of 5. “The overall user feedback is clear: the prototype is already usable and useful in the field as is. This is particularly promising, as there is still significant room for improvement,” emphasizes the founder of Neuralia. For Lionel May, these results confirm the solution’s potential. “Our technicians were pleasantly surprised by the relevance of the responses. Even though this is just the beginning, the tool is already providing them with concrete assistance in their daily work.”
Technical Challenges Overcome
The project was not without its challenges. “The main obstacles involved data quality and structure, the AI’s ability to interpret highly specialized data related to machine breakdowns, and choosing the right technical approach,” explains Gary Folli.
A key lesson from the project is that in AI, “you should never judge a project solely on initial assumptions. The only reliable approach is to test it in practice with a prototype and have it evaluated by domain experts,” he notes. “In this specific case, it wasn’t clear at the outset that AI would have such a significant impact in a manual line of work. “We tend to wrongly underestimate these use cases compared to administrative or cognitive roles.”
The Pivotal Role of The Ark
“The Ark played a key role throughout the project by providing financial support, which made it possible to launch the project despite initial uncertainties,” emphasizes Gary Folli. “But beyond the financial aspect, the strategic and technical advice provided was invaluable at every stage: features, product direction, and technical choices.”
For Lionel May, this support was crucial. “Without The Ark’s support, we probably wouldn’t have dared to embark on such a project. Their expertise allowed us to move forward with confidence.”
Toward a Comprehensive Platform: Maintrix
Buoyed by these encouraging results, the project doesn’t stop there. “The goal is to transform this prototype into a comprehensive platform, which will be called Maintrix,” announces Gary Folli. “Tests are currently underway at other companies facing the same diagnostic challenges, with equally encouraging results.”
At a Glance
Amount invested by The Ark Foundation: CHF 10,000
Project duration: 8 months (April 2025–December 2025)
Partners: Neuralia, STA, Cimark, The Ark