Leading omniscreen advertising platform.

Tasks

• Build machine learning (ML) pipelines for OTT screen classification
• Develop neural networks for TV channel recognition and detection
• Create image classification models for classifying video games on TV screens in real time

Challenges

Optimizing the algorithms while considering constraints and working with a highly complex set of CV and other algorithms. The solutions had to be executed on Smart TV platforms with significant MIPS constraints.

Solutions

• Built ML pipeline for OTT screen classification to classify streaming platforms (e.g., Netflix, HBO Max, Disney) on edge devices
• Developed neural networks for TV channel recognition and detection, used for recommendation systems
• Designed image classification models to classify video games (e.g., Call of Duty, Mario Series) in real time.

Outcomes

• Successful implementation of classification pipelines for OTT streaming platforms
• Developed accurate neural networks for TV analytics and viewer recommendations
• Minimal day-to-day task management required, enabling engineers to work independently with weekly syncs.

Technologies Used