At the NVIDIA GTC event in San Jose, Milestone Systems is set to highlight significant enhancements in its AI developer tools, particularly within the Hafnia platform.
The recent expansion includes the addition of Synthetic Data and an upcoming Training-as-a-Service (TaaS) feature, designed to equip developers to train AI models for not just familiar situations but also for unusual or previously unencountered scenarios.
Bridging the AI Gap
Hafnia aims to bridge the divide between data, training, and deployment by enabling the development of superior vision AI models. These enhancements allow the models to be deployed within Smart City environments, emphasizing proactive readiness rather than just reactive learning.
NVIDIA's Physical AI Data Factory Blueprint serves as the reference model for integrating data curation and augmentation with large-scale agentic orchestration, which facilitates the transformation of both real and synthetic data into comprehensive, physics-aware datasets available for model training.
Advancing Beyond Historical Data
Edward Mauser, Director of Hafnia at Milestone Systems, stressed the collaboration with NVIDIA. "Together with NVIDIA, we are taking Hafnia to the next level by combining trusted real-world data with synthetic augmentation," he stated, highlighting how this combination aids in training AI models that address both known and unforeseen scenarios.
AI typically learns from historical events, but unpredictable urban environments, including rare weather or traffic conditions, are often improperly represented in traditional datasets. Hafnia integrates synthetic data into its curated video library to address this, using NVIDIA's Cosmos Transfer to ensure rare situations and underrepresented classes are included, leading to a more comprehensive data balance.
Training-as-a-Service Unveiled
Milestone unveils TaaS at GTC, giving developers seamless access to Hafnia’s video dataMilestone is introducing its TaaS model at GTC, designed to consolidate fragmented training processes into a seamless solution, allowing developers to access Hafnia's comprehensive video data library for their training needs. This service enhances the ability to customize datasets and fine-tune models according to specific applications, ensuring compliance and traceability.
The streamlined process through TaaS is geared toward simplifying the training and management of data, allowing developers to build powerful video analytics solutions significantly faster.
VLM-as-a-Service for Smart Cities
In partnership with NVIDIA, Hafnia is offering VLM-as-a-Service, which includes a suite of Visual Language Models optimized for use in Smart City applications. The service includes EU-optimized models specifically for traffic, with additional models forthcoming to support a wider range of scenarios. These models aim to reduce costs related to data collection and infrastructure needs while showing marked improvements in several performance metrics.
The VLM-as-a-Service offers significant improvements, including a 19.4% increase in flow and direction accuracy, 8.9% better visual feature detection, and 4.4% enhanced alert verification accuracy, demonstrating its efficacy in real-world applications.
Comprehensive Cloud Infrastructure
Hafnia employs a flexible, multi-cloud strategy incorporating AWS, Nebius, among others, to support the AI model lifecycle. This approach particularly addresses data sovereignty issues, allowing full control and secure processing of sensitive data. The collaboration featuring the Milestone Synthetic Data Generation pipeline is set to launch on Nebius.
This infrastructure supports the development lifecycle comprehensively, from data sourcing and fine-tuning to extensive training and deployment, providing the required scalability and control.
Attendees at NVIDIA GTC can explore Hafnia’s expanded offerings through live demonstrations at Booth #2036. Edward Mauser will further elucidate on "A New Frontier for Vision AI with Expert Reasoning Agents" on March 18, aiming to unpack the advancing scope of VLMs for future applications.
At NVIDIA GTC in San Jose, Milestone Systems will showcase major advancements to its suite of AI developer tools coming out of Hafnia.
The latest expansion introduces Synthetic Data and a forthcoming Training-as-a-Service (TaaS) offering, enabling developers to train AI models not only for real-world conditions, but also for rare and previously unseen scenarios.
Bridging the gap
Hafnia bridges the gap between data, training, and deployment, allowing developers to reduce dataset bias while training best-in-class vision AI models and deploying them in Smart City solutions. The result: AI systems that move beyond reactive learning and toward proactive readiness across their entire lifecycle.
NVIDIA Physical AI Data Factory Blueprint is the reference architecture that unifies data curation, augmentation, evaluation, and agentic orchestration at scale. Powered by NVIDIA Cosmos™ open world foundation models and NVIDIA OSMO™, it transforms raw real-world and synthetic data into high-fidelity, physics-aware, model-ready training datasets—accelerating development with speed, scale, and reliability.
Training beyond historical data
“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.”
AI systems typically learn from past events. But real-world environments, like cities, are unpredictable. Rare weather conditions, unusual traffic patterns, or region-specific vehicle types are often underrepresented in traditional datasets.
Hafnia addresses this gap by integrating synthetic data into its curated, real-world video library. Synthetic augmentation through NVIDIA Cosmos Transfer allows developers to have access to data including rare or dangerous situations, balanced underrepresented object classes, model regional and environmental variations, and systematically reduced datasets biases.
Importantly, synthetic data doesn’t replace but enhances Hafnia’s real-world foundation. This ensures authenticity, compliance, and consistent annotation quality while expanding scenario coverage.
Training-as-a-service for the computer vision community
Milestone will preview its upcoming Training-as-a-Service at GTC. Instead of piecing together fragmented pipelines, developers now have a seamless bridge from Hafnia's data to robust training infrastructure, allowing them to focus entirely on building high-performing video analytics.
TaaS gives developers streamlined access to the compliant, high-quality video data within Hafnia’s library – both real-world and synthetic data sets - that can be used within their own training pipelines. They will be able to customize datasets and fine-tune models for specific use cases. Because the data within Hafnia’s library is compliantly sourced and fully traceable, developers can train models with the confidence their model training is compliant with relevant regulations.
By removing the complexity of sourcing and managing training data, Hafnia allows developers to focus on building high-performing analytics solutions up to 30 times faster.
VLM-as-a-service powered by NVIDIA
In partnership with NVIDIA, Milestones Hafnia also offers VLM-as-a-Service: a suite of Visual Language Models built on NVIDIA Cosmos Reason models and optimized for Smart City environments.
At GTC, Milestone announces the availability of a new EU-optimized VLM for traffic, already running with selected EU cities as customers. Additional models are coming soon, expanding the suite of VLMs offered through the VLM-as-a-Service platform to cover more smart city scenarios.
These hosted models break down the barriers to powering Computer Vision products with Generative AI solutions tailored for Smart Cities, eliminating costs for data collection, repeated retraining, and infrastructure scaling.
Performance evaluations show significant improvements over base models, including:
- +19.4% improvement in flow and direction correctness
- +8.9% improvement in visual feature detection
- +4.4% improvement in alert verification accuracy
For teams looking to accelerate the final stages of the lifecycle by 70 times, Hafnia’s VLM-as-a-Service provides optimized, ready-to-deploy Visual Language Models tailored for Smart City applications.
End-to-end cloud infrastructure
To support the complete vision model lifecycle, Hafnia is built on a flexible, multi-cloud strategy that leverages AWS, Nebius, and other providers. By uniting foundational cloud reliability with specialized AI computing, Hafnia ensures that every stage of development has the exact power it needs.
The Milestone Synthetic Data Generation pipeline using Cosmos Transfer and Cosmos Evaluator powered by Cosmos Reason is being launched on Nebius at GTC.
Importantly, this multi-cloud approach supports data sovereignty requirements, ensuring customers maintain full control over where their sensitive information is stored and processed.
From initial data sourcing and customized fine-tuning, to full-scale training and deployment, these collaborations provide the scalable power needed to manage the complete lifecycle of vision model development.
Meet Milestones Hafnia at NVIDIA GTC
At NVIDIA GTC, visitors can experience live demonstrations of Hafnia’s expanded data library, preview Training-as-a-Service, and explore the latest VLM capabilities. You can find us at Booth #2036, where the team will be showcasing these innovations throughout the event.
Edward Mauser, Director and Product Lead at Milestone Systems, will present “A New Frontier for Vision AI with Expert Reasoning Agents” on March 18, from 2:00–2:40 PM – exploring the next generation of VLMs.
Milestone is helping the computer vision community master the complete AI lifecycle, building smarter, fairer, and more resilient AI, ready not only for today’s environments, but for the unseen challenges of tomorrow.
