Team Sackriti won first place and a prize of NOK 40,000, while J2W took second place and received NOK 10,000. Congratulations to both teams!
At the heart of the competition was a real operational challenge identified by Soundsensing. The Norwegian technology company provides predictive-maintenance solutions for heating, ventilation and cooling systems in commercial buildings. Its battery-powered monitoring devices are deployed across many locations, making timely and efficient battery replacement an important part of keeping those systems running.
Soundsensing made data collected from real sensors in buildings across Norway available for the challenge. By opening this real-world dataset to participants, the company gave researchers, students and industry specialists an opportunity to develop and test ideas against a concrete predictive maintenance problem.
“Predictive maintenance is a multi-faceted problem that involves forecasting of asset health indicators, decision logic that accounts for uncertainty, and planning that respects practical constraints. By sharing this dataset, we wanted to see how new approaches could make predictive maintenance more efficient,” said Jon Nordby, Head of Data Science, Soundsensing.
Participants were asked to develop models that could estimate the remaining useful life of each battery. They then had to turn those estimates into practical replacement plans that accounted for travel time between buildings, the number of available workers and the limited length of a working day.
This combination made the competition more than a forecasting exercise. A successful solution had to produce accurate predictions while also supporting realistic operational decisions. The submitted solutions were reviewed with support from a scientific committee of researchers from NMBU and other universities.
“For me, every good competition is a well-designed game. Kudos to NORA.ai and Soundsensing for using Hugging Face to create a top-class competition. Every submission to the leaderboard was a thrilling experience, as we kept learning and optimising our chances. Until our very last submission, we weren’t sure we were going to make it. Overall, it was an incredibly enjoyable learning experience. I’m grateful for the opportunity and positively surprised by the outcome.”
— Sachin Gaur, on behalf of Team Sackriti
“Tackling real-world operational data with genuine physical constraints made BatterySwapAI an exceptional challenge. We're thrilled that our symbolic AI approach delivered robust results and performed so effectively against such a competitive field,” said Joël Wanza, Senior ASIC Architect at Numascale.
The competition attracted participants with backgrounds in data science, engineering, machine learning and applied research. Their different approaches demonstrated how industry data can inspire a wide range of ideas—from time-series forecasting and battery-health modelling to optimisation and work-order planning.
“The challenge brought together talented people from research and industry around a concrete problem. The winning solutions demonstrate the value of open data and collaboration,” said Kushtrim Visoka, Senior Advisor at NORA.ai.
The BatterySwapAI 2026 Challenge brought together NORA’s research and innovation network, Soundsensing’s industry experience and real-world data, and scientific evaluation supported by researchers from NMBU and other universities. The competition offered prizes totalling NOK 50,000 and encouraged participants to make their solutions open and reproducible.
NORA and Soundsensing thank everyone who participated, along with the scientific committee and all those who helped make the challenge possible. Congratulations to Team Sackriti and J2W on their achievements.