This October, Glencore took part in the 10th edition of the InCube Challenge in Zurich, a competition organised by the ETH Entrepreneur Club that brings together students to solve real-world business and technology challenges.
As a sponsor of one of the event’s cubes, Glencore’s Environment team set five students a challenge: How could we use AI-assisted coding to build and connect multi-part water models faster and more reliably?
Over five days, the students, known as "Cubees", lived and worked together in the Glencore cube, developing their solution and presented their answer at the final pitch in the Audi Max at ETH Zurich on 30 October, where they placed second overall.
The Glencore Cubees outside the InCube cube at ETH Zurich
The challenge
Water is an increasingly precious resource that it is essential for society, nature and many socio-economic activities. Glencore uses water modelling to inform our approach to water stewardship and operational decision making. Modelling helps our industrial assets manage their water use responsibly, manage and mitigate their environmental impacts, plan for droughts and extreme weather events, and support safe, efficient water use throughout the industrial asset lifecycle.
Water models are also fundamental to the responsible development of new mining projects. They underpin project design, permitting, environmental controls, community confidence and lender assurance, and they shape how we understand and manage our impact on the environment. Faster, more transparent modelling could lead to earlier risk identification, better water-management decisions and increase the efficiency of project assessments.
But producing good models is no easy task. Environment teams across our operations have to build and integrate various models, like climate, chemistry and water-balance models. Each step, from climate and hydrology to groundwater, geochemistry and water balance, is run using specialist software. Bringing the outputs together involves manually transferring data between different modelling tools. This means that when water models are re-run, a loss of time is spent on transferring results from one modelling component to the next, rather than on the scientific analysis itself.
The Cubees were provided with an illustrative water model of a fictional site using real world open-source data where acidic groundwater contaminated by historic mining activities has travelled around 25 kilometres. This fictional site’s "acid plume" included monitored data since 1984, providing a long-term dataset for modelling.
Their task was to run the illustrative water model from start to finish, test 23 different scenarios and develop automated quality checks that could spot errors before they affected the results. The students were encouraged to use AI coding assistants to reduce the time needed to build and connect the different parts of the model, while also showing that these efficiencies could be achieved without losing the technical quality, traceability and expert oversight that regulators and communities rely on.
The solution: Aval
The Cubees' answer is Aval, a data collaboration platform that sits between each of the components of a water model. Teams first agree what each dataset that makes up the model must contain. Aval then checks every file on upload, suggests fixes that the data owner approves, and alerts only the teams affected. The result: weeks saved for environment teams and less human errors.
“The Cubees demonstrated that there is real potential to streamline the way complex environmental models are built and connected without compromising quality or oversight. While the challenge was rooted in a water-modelling problem for the mining and metals sector, the solution could have applications across many industries that depend on water modelling to manage risk and make informed decisions.”
Jason Jones - Glencore Group Environment Manager
In the students' words:
“What I liked most about the challenge was that curiosity always drove us. We didn't stop at the surface, but dug deeper to find the real problem and its cause. When we presented our solution at the end of the week, experts said it would have saved them a lot of time, and I was genuinely proud of what we had built.”
Andri - Cubee
“I wanted to push my limits, but building Aval so fast ultimately came down to the people I was working with. InCube proved that the absolute best way to build something real is to surround yourself with teammates who constantly set the bar higher for you.”
Sofia - Cubee
Investing in innovation and talent
Investing in innovation and talent
Taking part in the InCube Challenge lets Glencore engage with talented students across a range of disciplines, strengthen our links with leading universities and research communities, and gain fresh perspectives on complex business problems.
A strong prototype could also help inform how we pilot AI-assisted model integration on a live project.
Programmes like this show the value of bringing industry expertise together with emerging talent. We thank our Cubees for their hard work, the ETH Entrepreneur Club for hosting, and everyone at Glencore who helped shape the challenge and share their expertise.