Targa Telematics rolls out on-premise AI for developers

Targa Telematics has completed and deployed an on-premise generative AI software development environment, now used by about 150 engineers across the company.

Delivered on Hewlett Packard Enterprise infrastructure in collaboration with Gruppo Zenit, the platform is built around large language models and open-weight models. It is intended to support software development while keeping code, data and model use within Targa Telematics' own infrastructure.

In the first seven weeks after deployment, the platform handled almost 30,000 requests from developers, processed more than 3.4 billion tokens and generated 65 million tokens in response, reflecting early take-up by engineering teams.

The project is centred on an on-premise approach rather than public cloud or third-party inference services. This means proprietary information does not need to be transferred outside the business when engineers use the tools for coding support or as an AI assistant.

The move comes as businesses face growing pressure to balance wider use of generative AI with requirements around data handling, cybersecurity and compliance. For companies working with internal source code and other sensitive information, where data is processed has become a key factor in deployment decisions.

Open-weight models form part of that calculation because they can be adapted and run in-house. This allows Targa Telematics to benefit from the pace of development in the open-source ecosystem while retaining oversight of model governance and customisation for internal use.

Carlo Stefanelli, CTO of Targa Telematics, outlined the company's rationale for the project. "Generative AI is increasingly becoming a day-to-day tool for our developers. We wanted to create a fully on-premise development environment while retaining complete control over our code, data and technological know-how," he said.

"An on-premise approach, based on open-weight models, is key to ensuring that all information remains entirely within our corporate environment, allowing us to foster innovation without compromising on data security and control."

Infrastructure focus

Gruppo Zenit worked on the design of the underlying infrastructure, with an emphasis on security, business continuity and long-term data governance. Hewlett Packard Enterprise provided the computing environment for AI inference as well as management tools for the system.

The technology stack was designed to support reliability, scalability and secure administration within the company's own environment. That reflects a broader trend among businesses seeking to test and expand generative AI tools without moving sensitive datasets into external infrastructure.

Giuseppe Barbera, CEO of Gruppo Zenit, said the project showed how AI use can move from isolated testing to a more established internal platform. "AI can only generate value when the underlying infrastructure is designed to be secure, scalable and manageable. Our role is to support companies in developing ecosystems where innovation and data protection can grow and flourish hand in hand. The project we have carried out with Targa Telematics shows how AI experimentation can be transformed into a platform primed to drive business evolution."

Targa Telematics develops digital solutions and platforms for connected mobility and operates across several European markets as well as Chile. The new setup strengthens its internal experience in generative AI and provides an infrastructure model that could be reused for other applications.

That gives the project significance beyond developer assistance alone, providing a test case for how a company with sensitive operational and software data can deploy large language models within its own environment.