Advancing Sustainable Materials Intelligence with Naia

We’re excited to spotlight the next story in our series on how EuroCC Denmark helps Danish innovators push the boundaries of what’s possible with HPC. 

This month, we turn to Aarhus-based AI startup Naia, a company building domain specific language models for industrial sustainability and cost estimation.

The challenge 

General-purpose LLMs struggle with highly specialized industrial data. Naia needed a model capable of identifying materials, interpreting technical properties, and matching bill of materials (BOM) components to sustainability and cost data — even when inputs were incomplete, inconsistent, or noisy. Achieving this required domain-adaptive training at a scale far beyond local compute. 

The solution 

Through EuroCC Denmark, Naia gained access to EuroHPC's MareNostrum5 supercomputer and hands-on guidance for scaling their AI workflows to HPC infrastructure. Our team supported Naia in preparing their DeepSpeed and PyTorch pipeline for the transition, ensuring they are well-positioned to fine-tune a compact, efficient LLM on curated and synthetic materials science datasets. With the infrastructure and expertise in place, Naia has the foundation to scale training runs efficiently and iterate rapidly as their data and model requirements grow. 

The impact 

Naia’s new model empowers industry partners to integrate sustainability insights directly into early stage material selection. This accelerates compliance, reduces environmental impact, and strengthens decision making across manufacturing and supply chain processes. Additional benefits include: 

• Higher accuracy in BOM to sustainability matching  

• Private deployment with full domain control 

• Energy efficient HPC training with a smaller carbon footprint  

At EuroCC Denmark, we’re proud to help startups like Naia transform complex industrial challenges into scalable, sustainable AI solutions.

More cases are on the way,  stay tuned for the next story in our series.