Velvet

You'll **build ML pipelines** that turn raw footage into clean training data for frontier multimodal AI models. Your work directly powers **data realism** for the next generation of AI. You'll own the full lifecycle from script to scale across thousands of hours of video.

You'll own data procurement and post-processing pipelines for frontier model training data. Your work directly impacts the quality of audiovisual datasets used by leading AI labs. This role combines operations, scripting, and client interaction for a high-impact startup.

You'll **develop and fine-tune models** for video and audio enhancement, conducting **data-oriented research** to push multimodal quality. You'll collaborate with engineers to integrate models into **large-scale processing workflows**.