Challenges in AI model updates after data annotation at scale and speed

Let’s say i have CCTV cameras installed in warehouses across multiple geographies and i need to build upon the computer vision ai use cases such as forklift detection, people safety, workforce efficiency etc.

In such a case, As an AI head/mlops i would prefer to use mature data annotation tools and services such as Labelbox or Labellerr or LabelGPT but the challenge will be rest of the process either the model training or deployment etc and model updates etc.?

Anyone used Nvidia stack in such a setup and what were the challenges faced across the complete workflow? My prospective customers are looking at the integrations where they would use Labellerr for their data annotation part of the workflow.