Scale AI · Staff software engineer
AI-assisted video labeling
I built a replacement for a full-motion video labeling application and integrated on-demand segmentation. A contributor can use the model to outline an object while working through the video. I also work on auto-labeling to take repetitive steps out of that workflow.
The engineering question I keep coming back to is where a model saves the contributor work and where the interface should make it easy to review or correct its output.