AI, data and robotics: How Everest Labs is transforming recycling against hard labor costs

11 Mar 2024 --- We explore the world of recycling powered by robotics and AI in conversation with Everest Labs CEO and founder JD Ambati. The company’s RecycleOS software is designed to identify PET plastic into three specialized streams and robotically sort the materials into three different bunkers.

Everest Labs’ process is said to be “unique” as most facilities that recycle PET, collect and bale the three streams as one. “The ability to filter these plastics into PET bottles, thermoforms and pigmented PET marks a significant step in our commitment to innovation and environmental stewardship in our community,” says the company.

Everest Labs recently received a grant from The Recycling Partnership Funds in partnership with Caglia Environmental for its AI and robotic technology deployment at Cedar Avenue Recycling & Transfer Station material recovery facility in California, US. 

How does Everest Labs address labor challenges in the recycling industry?
Ambati: Keeping recycling plants fully staffed and productive is a real struggle for plant operators. There are high turnover rates, soaring labor costs (especially with contract labor), absenteeism and safety concerns. As a result, quality control suffers, and too often, what could be recycled is landfilled. EverestLabs offers a straightforward, cost-effective solution to help plant operators solve these labor challenges. 

Recycling facilities can place RecycleOS, our AI and robotics solution, on lines that they can’t staff or even where humans aren’t able to sort. RecycleOS can also work alongside human sorters, helping improve recovery and efficiency. ACI, a customer based in the San Francisco Bay Area, saw a 59% reduction in labor costs when using EverestLabs’ AI and robotics. Packaging manufacturers can use RecycleOS to identify and recover brand-specific packaging and assess different packaging types’ recycling and recovery rates.

While our robots require an upfront expense, most multi-shift recycling plants see a payback in less than six months. This payback is against hard labor costs alone and doesn’t include the savings facilities see from increased material recovery, decreased landfill fees, turnover rates, lower training costs and reduced injuries. 

For larger enterprises who want to upskill workers, robotics is an easy way for them to perform dull and dirty jobs.

Everest Labs offers “Robot as a Service” designed for ease of use and eliminates the burden for MRF operators. The only thing required of MRF employees is simple, regular maintenance of the robot’s end-of-arm tools. Our simple-to-maintain design requires just one suction cup change weekly and occasional end-of-arm cleaning — both of which can be completed in a matter of minutes. We provide our customers with a short training and a packet on how to do this. Everest Labs also provides consumables, repairs and replacements at no additional cost. 

As part of our service, we provide 24/7 monitoring of RecycleOS with our Robot Operations Center (ROC). This monitoring ensures the performance and health of our systems and immediately contacts our team and the MRF if any problems need to be addressed, taking the burden off of the MRF operator. 

How does RecycleOS-powered robotic sorting technology address traditional PET recycling limitations?
Ambati: According to the National Association for PET Container Resources (NAPCOR)’s 2020 PET Recycling Report, PET plastic bottles have a recycling rate of only 27% (US). Other types of PET have an even lower rate of only 9%. The PET that does end up in recycling bins is sometimes lost at MRFs due to sortation challenges. 

In fact, at the recycling facility level, about 30% of all recycling materials sent to these facilities end up in landfills. This is where Everest Labs comes in. We can help increase PET recovery at the MRF level by improving facilities’ operational efficiency and material recovery. 

RecycleOS is made up of AI, data and robotics. RecycleOS AI powers both our robotics and data. Our robotic system occupies the same space as a human sorter, but it improves recovery rates 2-3x, with an uptime of over 99%. We also provide our customers with AI-powered data on material composition, bale composition, and value lost to landfill. This data helps our customers tune their upstream equipment, like optical sorters. Combining our robotics and data ensures maximum efficiency and recovery throughout a facility.

What quality control mechanisms does Everest Labs have in place to guarantee accuracy across recycling facilities?
Ambati: Our advanced neural network is trained on a proprietary dataset of 4B+ recyclable objects and is updated continually for new classes of materials and varying belt conditions. Our AI can sort objects into more than 50 classes of recyclable objects with more than 95% accuracy. This allows us to accurately distinguish between the different types of PET plastics for recovery. 

To guarantee the accuracy and consistency of RecycleOS across various recycling facilities, we provide closed-loop monitoring of all our robotics using sensors-based audits that track AI and robotics performance, as well as our globally staffed ROC, who oversee the robotics fleet. The ROC provides 24/7 monitoring of robot, AI, and plant performance and health for every RecycleOS system. Everest Labs advanced neural network is trained on a proprietary dataset of 4B+ recyclable objects.

Could you elaborate on the scalability of RecycleOS technology and applicability to other materials?
Ambati: Everest Labs can transform every step of the packaging supply chain. In MRFs, RecycleOS can automate sorting on residue, QC, and baler clean-up lines. Everest Labs is able to identify and recover 50+ classes of recyclable materials and contamination. Our solution is also deployed in Caglia Environmental’s last chance line, focusing on aluminum can recovery.

RecycleOS is also used in reclaimers, otherwise known as secondary processing facilities. Reclaimers face challenges sourcing and verifying high-quality bales of recycled content. RecycleOS helps by ensuring they receive high-quality feedstock from their suppliers and reducing contamination in their waste stream to ensure high yields of outgoing materials and reduced downtime. 

Packaging manufacturers can use RecycleOS to identify and recover brand-specific packaging and use RecycleOS data to assess different packaging types’ recycling and recovery rates. This data can be used to optimize packaging design and evaluate the effectiveness of recycling programs. We can also provide companies with auditable reports on sustainability metrics that can inform EPR and ESG initiatives. 

Do you foresee any impacts on the packaging industry with the incoming EU AI Act?
Ambati: RecycleOS AI uses advanced computer vision and deep learning to identify packaging attributes and other objects in the recycling stream. The use case is not considered high risk by the EU AI Act. The models are not trained on copyrighted or personally identifiable data that could put people or companies at risk. The output of the AI model is only used to quantify the performance of recycling/sorting processes and optimize/automate these steps. Therefore, upcoming legislation should not negatively impact the benefits that AI will bring to the circular economy.

By Radhika Sikaria