The Most Ambitious Company You’ve Never Heard Of Is Redefining Trust in AI

When we talk about the global AI race, the conversation is dominated by the United States and China. Despite having innovators like Mistral and Hugging Face, Europe rarely features in that narrative. Its rigorous stance on data privacy, user safety, and regulatory oversight is often portrayed as a drag on innovation rather than a foundation for responsible growth.

Domyn — formerly known as iGenius — wants to change that perception. The Milan-based company is positioning itself as Europe’s sovereign AI pioneer, determined to prove that world-class innovation can coexist with European values of governance and trust. With plans to build AI gigafactories powered by NVIDIA’s latest Blackwell chips, Domyn aims to provide regulated industries with safe, explainable, and high-performance AI systems — a credible European alternative to U.S. giants.

It offers two foundational language models: a 355-billion-parameter LLM for mission-critical use cases and a 10-billion-parameter model for lightweight enterprise tasks. But it wants to do a lot more.

From its headquarters in Milan and offices in New York, Domyn is building AI tailored to the nuanced demands of sectors like finance, healthcare, and government. Recently, it announced the launch of its Global Financial Services Division, a sovereign AI research unit designed to transform how financial institutions adopt and deploy artificial intelligence across mission-critical businesses.

To strengthen its financial expertise, the company recently appointed former BlackRock managing director Stefano Pasquali to lead its newly launched financial services division.

Founded by Uljan Sharka, Domyn is betting big on a future where compute power and responsible AI design define competitiveness. Yet few companies outside Silicon Valley have the resources or ambition to match such aspirations. While OpenAI secures billion-dollar GPU and cloud deals to power its frontier models, Domyn is carving a different path — one rooted in European sovereignty, domain specialisation, and transparency.

Domyn also has a secret weapon at its disposal– one of the largest supercomputers in Europe. “Training foundation models from scratch, especially trillion-parameter LLMs, requires not just scale of computation, but also infrastructure ownership. 

That ownership is critical when you’re building responsible AI for regulated industries: you need transparency, security, and full traceability into why a model made certain decisions,” Bhaskarjit Sarmah, Head of AI Research for Financial Services at Domyn, told The Left Shift. 

Building Foundational Models in Europe

Like Pasquali, Sarmah also joined Domym from BlackRock. “From the outset, Domyn has focused on making AI accessible to all sectors, enabling them to adopt AI with confidence, even those with strict compliance requirements, such as financial services, government, and advanced manufacturing.

“How do we do that? By delivering AI solutions that enterprises can truly own, govern, and trust. In other words, we make it our mission to help organisations achieve full sovereignty over their proprietary intelligence — including data, models, and insights,” Sarmah said.

More broadly, Domyn’s vision is to advance responsible intelligence that elevates humankind. The company views AI as a positive force capable of amplifying human potential and tackling complex, real-world challenges — particularly within the financial services sector.

“To make AI accessible in such highly regulated industries, we commit to developing solutions that are ethical, responsible, and rooted in human values. Ultimately, our goal is to broaden access to technology and empower financial services organisations to use it with complete confidence and safety,” Sarmah said.

Moreover, Domyn's models come with an open-source licence. Since the models are from vendor lock-in, enterprises are able to modify, train, and distribute the models as they choose.

"In line with Domyn’s mission, an open-source license makes it possible for enterprises to get full sovereignty over a model while guaranteeing complete transparency. We believe that models like Domyn’s should be open, and therefore traceable and auditable so that every output can be inspected and verified. This stands in direct contrast with opaque, black-box systems that obscure logic, data provenance, and security components, ultimately going against our quest to provide ethical, responsible AI for all," Sarmah added.

The Power of Europe’s Largest AI Supercomputer at Its Disposal 

In the rolling hills of southern Italy, Domyn is constructing Colosseum, Europe's largest AI supercomputer. This monumental project, developed in partnership with the UAE’s G42 and NVIDIA, among others, is being developed at a cost of $1 billion.

Originally built to deliver 115 exaFLOPS of computational power using renewable Italian energy, the system will integrate NVIDIA DGX GB200 units equipped with thousands of Grace Blackwell Superchips, enabling real-time training and inference of trillion-parameter models.

 “Colosseum doesn’t just let us build big models, it gives us the ability to build them right for regulated industries like finance: transparent, auditable, secure, and custom-tailored. That’s a class of research breakthroughs both technical (architectures, interpretability) and operational (deployment, governance) that simply wouldn’t be possible under more limited compute or third-party infrastructure,” Sarmah pointed out.

Competing with the Likes of OpenAI

Domyn’s desire to build foundational models for enterprises puts it in direct competition with the likes of Google, Microsoft, and OpenAI. Most enterprises are developing pilots by leveraging AI models from these companies, including many financial sector companies. 

Sarmah, however, claims Domyn’s focus is different. “We are only focussed on building Responsible AI for Regulated Industries (like Finance, Healthcare etc.), whereas OpenAI is very generic in nature. OpenAI does not care about Responsible AI. 

“People might use OpenAI for low-risk use cases like productivity gain use cases. But when it comes to high-risk use cases (mission-critical use cases), customers will come to us because we are building Responsible AI, which can help you to detect and mitigate risks.”

Looking ahead, Domyn plans to expand its model portfolio and develop AI agentic capabilities that can autonomously support complex enterprise workflows. These innovations aim to combine automation, compliance, and explainability, enabling businesses to deploy AI responsibly while maintaining full control over critical decisions.