The success of artificial intelligence hinges on access to high-quality, annotated data, yet developers struggle to find sufficient sources for training visual AI models. To address this challenge, Milestone Systems has announced Project Hafnia, a new platform designed to provide compliant and traceable video data, leveraging NVIDIA’s Cosmos Curator and fine-tuning microservices.
Project Hafnia aims to accelerate AI development by aggregating and curating video data from global partners and customers. The initiative represents a shift towards democratising AI model training, ensuring developers can access regulatory-compliant datasets that meet the growing demand for precision and accountability in AI applications.
Milestone’s focus is on enabling AI developers to access Vision Language Models (VLMs) fine-tuned using NVIDIA’s platform, with an initial emphasis on traffic management. The first service offering will provide a VLM optimised for transportation analytics, designed to run efficiently on NVIDIA GPUs and within NVIDIA’s video search and summarisation (VSS) AI blueprint.
Deepu Talla, Vice President and General Manager of Embedded and Edge Computing at NVIDIA, believes this approach will help unlock the next phase of AI evolution. “By leveraging the NVIDIA platform, Milestone Systems is helping accelerate this next wave of powerful visual services,” he said. “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.”
Bridging the AI data gap
Milestone Systems is tapping into its extensive network of partners to compile a comprehensive and compliant video dataset. The platform is designed to remove major barriers in AI training by ensuring the data is properly annotated and meets stringent regulatory standards.
“Artificial intelligence is our generation’s biggest game-changer,” Thomas Jensen, CEO of Milestone Systems, said. “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 training AI models.”
By integrating NVIDIA Cosmos Curator’s data curation capabilities, Milestone claims the new platform could speed up AI model development by up to 30 times compared to current methods. This efficiency gain could have far-reaching implications for sectors such as manufacturing, airports, law enforcement, and smart city infrastructure.
Enhancing AI model accuracy
The first two services under Project Hafnia will focus on AI model training and traffic management. A Training-as-a-Service model will allow developers to access curated, high-quality video data to improve the accuracy of AI models, while a VLM as a Service will provide an industry-leading solution for smart city transportation and intelligent traffic services.
Milestone’s transportation-focused VLM is set to support a broad range of applications, including general traffic assessments, driving condition evaluations, alert validation, and incident reporting. With more precise data and advanced annotation capabilities, AI-powered analytics could help optimise road networks, reduce congestion, and enhance urban mobility.
As the platform launches as a pilot programme, Milestone is inviting developers to sign up for early access, initially focusing on traffic-related video data. The long-term goal is to expand its scope to cover additional sectors, ensuring AI developers have the resources they need to build high-performance models across multiple industries.
The project signals a significant step towards improving AI’s reliability by ensuring training data is accurate, transparent, and accessible. If successful, it could redefine how AI models are trained and deployed, setting a new benchmark for data-driven AI development.
