EverestLabs.ai is a pioneering climate technology company committed to revolutionizing the recycling industry through the integration of artificial intelligence (AI) and automation. Their mission is to decarbonize the planet by ensuring all recyclables are efficiently recovered, thereby promoting a sustainable circular economy.
At the heart of EverestLabs’ offerings is RecycleOS®, the world’s first AI-powered operating system designed specifically for recycling facilities. This comprehensive platform provides a suite of solutions aimed at enhancing the efficiency and effectiveness of Material Recovery Facilities (MRFs):
- Data Solutions:
- Material Characterization: Enables facilities to understand plant performance by capturing object-by-object data at any point in the recycling stream.
- Plant Equipment Optimization: Assists in fine-tuning machinery for optimal performance.
- Bale Tracing & Audits: Facilitates tracking and verification of processed materials.
- Municipal and EPR Compliance: Ensures adherence to local regulations and Extended Producer Responsibility mandates.
- Robotics Solutions:
- RecycleOS® integrates advanced robotic cells equipped with AI-driven vision systems to automate the sorting process, thereby increasing accuracy and reducing reliance on manual labor.
To support seamless integration and operation, EverestLabs offers an all-inclusive service package that encompasses hardware, software, and professional services. This includes 3D vision systems, robotic arms, on-premise compute modules, RecycleOS® software, customized data analytics, continuous AI model updates, deployment services, and 24/7 monitoring through their Robotics Operations Center (ROC).
AI Models and Technological Innovations
EverestLabs has developed proprietary AI models that are central to the functionality of RecycleOS®. These models are trained to accurately identify and sort a wide array of recyclable materials, even when items are crushed, soiled, or partially obscured. The AI system processes images from 3D cameras at 30 frames per second, achieving material classification in approximately 8 to 12 milliseconds. This rapid processing enables the robotic arms to operate with a picking accuracy exceeding 90%.
The AI infrastructure includes a custom-built machine learning framework featuring:
- A web-based annotation tool for data labeling and auditing.
- A scalable data ingestion system capable of managing diverse datasets.
- A rapid data generation mechanism with caching for efficient training over large datasets.
- A model training pipeline compatible with multiple machine learning frameworks.
- A deployment system for seamless integration into production environments.