Most computer vision models work in controlled environments. Very few work reliably across different cameras, locations, and real-world conditions.
In this first Hafnia Tech Talks – a new series by Milestone Systems – we heard from Katja Sirazitdinova, Senior Developer Advocate at NVIDIA, on vision-language models and video analytics AI agents — and how they’re helping developers move from development to deployment of computer vision solutions in complex city environments.
In this session, we covered:
The gap between what AI can do in the lab and what cities need in the field
The challenges of scaling across real environments (different cameras, data quality, conditions) – and how developers are solving them
A look at NVIDIA's agentic blueprints and models, via Hafnia, and how developers can use them to build for city-scale complexity
Where the real value lies – moving beyond detection to decisions that improve operations, safety, mobility, and services.
This was an interactive session moderated by our Community Lead, Roland Harwood.
