The video recommendation engine became STRV’s biggest ML project to date. With no data available, we created a custom dataset and built an AI solution that relies solely on a highly advanced algorithm. We then adapted the AI to allow searching for videos by passing full-text queries.
Before Cinnamon approached STRV, the company had also acquired an ML solution for emotion recognition. Although the team decided to omit using it due to privacy concerns, we suggested that — with the right improvements — Cinnamon could offer the solution as a standalone B2B product.
STRV recognized that the ML model’s size would require very expensive infrastructure, and efficiency was lacking. We built an entirely new embedded AI that runs on clients’ devices, which saves on infrastructure costs, scales naturally and is remarkably privacy-conscious. Our solution ended up being 99% more efficient and 40% more accurate than the initial one — making it valuable to content creators such as Hollywood film studios.
