AI developers worldwide face a common problem—finding high-quality, reliable video data to train visual AI models. To tackle this challenge, Milestone Systems has announced Project Hafnia, a new initiative aimed at democratising AI model training by providing access to high-quality, compliant video data. The project will utilise NVIDIA Cosmos Curator and AI model fine-tuning microservices to streamline the process.
By leveraging NVIDIA’s platform, Project Hafnia will serve as a hub for both data generators—who can share and utilise their data—and developers, who will gain access to traceable and regulatory-compliant annotated video data. The ultimate goal is to put the latest Vision Language Models (VLM) and data curation technologies into the hands of as many developers as possible.
One of the first offerings from Project Hafnia is a fine-tuned VLM, optimised using NVIDIA’s fine-tuning microservices. This model has been trained on a vast dataset of compliant transportation video footage, curated with the help of NVIDIA Cosmos Curator. The VLM is designed for high accuracy and performance, optimised for NVIDIA GPUs and NVIDIA’s Video Search and Summarisation (VSS) AI blueprint.
“By leveraging the NVIDIA platform, Milestone Systems is helping accelerate this next wave of powerful visual services,” says Deepu Talla, Vice President and General Manager of Embedded and Edge Computing at NVIDIA. “The next phase in development and adoption of visually perceptive Agentic AI services will be unlocked by recipes like NVIDIA VSS blueprint combined with widely available and accessible fine-tuned VLM models.”
Expanding access to compliant video data
Milestone Systems, a leader in data-driven video technology and analytics, is bringing together compliant video data through its extensive global partner and customer network. These organisations, which specialise in video data management, can leverage their own video footage to develop more advanced analytics.
“Artificial intelligence is our generation’s biggest game-changer,” says Thomas Jensen, CEO of Milestone Systems. “A major challenge for ongoing development is having access to enough high-quality data for training AI models. The Project Hafnia platform will collect and curate data with the aspiration to be the world’s smartest, fastest, and most responsible platform for video data and AI model training.”
Project Hafnia is designed to remove major obstacles in the AI training process by providing a seamless method for accessing high-quality, curated data. Powered by NVIDIA Cosmos Curator, the platform is expected to accelerate AI and analytics development by up to 30 times faster than current industry standards.
AI training services and transportation focus
Milestone Systems will offer two core services under Project Hafnia:
- Training as a Service – AI developers can access high-quality video data to train their AI models efficiently.
- Visual Language Model as a Service – An industry-leading VLM designed for smart city transportation and Intelligent Traffic Service applications.
With improved data quality and enhanced annotation techniques, AI-driven analytics software could soon enable large-scale automation across industries such as traffic management, manufacturing, airports, law enforcement, and business operations.
By integrating NVIDIA’s platform, Milestone Systems has also taken a significant step in building a traffic and transportation-focused Visual Language Model. This specialised AI model can be used for general traffic assessments, driving condition evaluations, alert validation, and incident reporting.
Currently, Project Hafnia is launching as a pilot programme, and developers can sign up for early access at hafnia.milestonesys.com/joinwaitlist. The initial phase will focus on traffic-related video data, with plans to expand once the platform becomes fully operational. Milestone Systems will own and manage the platform, ensuring it remains a secure and efficient tool for AI developers worldwide.
