Milestone Hafnia Launches Synthetic Data Training Tools
Expanded AI suite at NVIDIA GTC introduces Training-as-a-Service to help developers prepare for rare and unpredictable security scenarios.
- By Jesse Jacobs
- Mar 17, 2026
Milestone Systems announced major advancements to its Hafnia AI developer tools at the NVIDIA GTC conference, introducing synthetic data and a forthcoming Training-as-a-Service (TaaS) offering.
The expansion allows developers to train vision AI models for both real-world conditions and rare scenarios that are often missing from traditional datasets. Using the NVIDIA Physical AI Data Factory Blueprint, the platform transforms raw and synthetic data into physics-aware training sets.
"Together with NVIDIA, we are taking Hafnia to the next level by combining trusted real-world data with synthetic augmentation," said Edward Mauser, director of Hafnia at Milestone Systems. "This enables developers to train AI models that are not only accurate in known situations, but also resilient in the unexpected."
Traditional AI learns from historical data, which often lacks examples of unusual weather, rare traffic patterns, or specific regional vehicle types. Hafnia addresses this by integrating synthetic data through NVIDIA Cosmos Transfer to reduce dataset bias and model environmental variations.
The upcoming TaaS feature aims to provide a bridge between Hafnia’s data library and training infrastructure. Milestone claims the service will allow developers to build analytics solutions up to 30 times faster by removing the complexity of sourcing and managing compliant video data.
Additionally, Milestone announced the availability of a new European Union-optimized Visual Language Model (VLM) for traffic. The model is part of a VLM-as-a-Service suite built on NVIDIA Cosmos Reason models. Performance evaluations of these models showed a 19.4% improvement in flow and direction correctness and an 8.9% increase in visual feature detection compared to base models.
To support the infrastructure, Hafnia utilizes a multi-cloud strategy involving AWS and Nebius. This approach is designed to meet data sovereignty requirements, allowing customers to maintain control over where sensitive information is processed.