From insights to action: Inside the new PolyPerception AI-native platform
Recyclers, MRF and PRF operators are dealing with changing regulations, unpredictable waste streams and too much data. To help sorting plants handle these challenges, PolyPerception recently launched a new AI-native platform. This is the first major release since TOMRA took a 51% majority stake in the company.
We spoke with Parshva Mehta, COO and Co-Founder at PolyPerception, about optimizing plant operations by adding an AI layer.
How is the platform transforming day-to-day operations for sorting plants?
For any high-throughput plant, the main priorities are always the same: keeping output purity high, minimizing valuable losses in the residue stream and ensuring the line keeps running. Our new platform acts as a virtual assistant, delivering expert insights into plants and their processes to support smarter decisions. Instead of relying on spreadsheets for answers, operators can ask questions in plain language and get clear answers in real time. It is the next step for our analyzer technology, which uses cameras along the line to track material from start to finish.
How does the new platform differ from other solutions available on the market today?
Most standard tools provide data that can feel overwhelming. Our AI platform is designed to be accessible, ensuring that operators can gain actionable insights from their own plant without needing to be data scientists. By understanding a customer's specific sorting processes, compliance targets and financial goals, the platform can instantly analyze production data to generate bespoke quality reports and configure relevant alerts. Being able to access insights like this is a first for this industry.
Why is the ability to interact with plant data by ‘chatting’ to the AI platform a breakthrough for a plant manager?
Historically, obtaining necessary data has required sifting through dashboards or relying on external data specialists. Now, an operator can query the system by asking, for example: ‘How did the recent adjustments to the recovery line impact our purity?’ The platform uses the necessary context to provide a direct, data-backed answer within seconds.
For instance, the platform helped a MRF investigating the quality of its polypropylene and polyethylene streams. Staff noticed an accumulation of bottles in the tray stream toward the end of the week. When they queried the platform, they found that contamination levels had been steadily rising throughout the week. While the platform flagged the issue within 45 minutes of it starting, these kinds of gradual shifts can be difficult to track manually with the high volumes moving through a plant. The platform provides the visibility needed to spot and fix these subtle trends before they become a bigger problem.
How does the platform stay accurate when the infeed material changes?
Inbound waste varies significantly according to seasonal cycles, annual trends and even the weather. At the same time, the industry is experiencing a surge in complex waste types, such as vapes, electronic cigarettes and batteries, all of which pose significant fire risks.
To help facilities respond to these evolving streams, we have introduced two sophisticated search methods. With similarity search, operators can identify a problematic object, such as an electronic vape, to instantly locate every other visually similar item in the stream. This means operators can spot fire hazards, like batteries, instantly without ever needing to retrain an AI model. We also include automated alerts for these high-risk items the moment a threat is detected.
Additionally, text and brand search allows users to filter for specific manufacturers or object types, such as 'filled refuse bags' or 'diapers', to monitor exactly what is passing through the facility in real time. We also collect data continuously during production, which enables us to update our deep learning models every two weeks to ensure high precision.
How do you help customers move from their existing process to this new way of working?
Trust takes time. These facilities operate at high throughput, so any mistake or line shutdown can mean a big financial loss. Understandably, operators are cautious about changing a process they know works.
We focus first on helping them understand and validate the data rather than asking them to change their behavior immediately. We work alongside them, compare our insights with their existing process, and demonstrate where value is being lost in the residual stream each week. Once they can see that cost in real terms and verify the data themselves, trust develops naturally. From there, adoption becomes much easier because they see the value firsthand, and many customers begin realizing a return on investment within the first month.
How much maintenance does the platform require?
It is designed for high-uptime environments and requires minimal maintenance. We provide all ongoing support as part of the contract, and we regularly check in with customers to ensure the dashboards, reports and alerts remain aligned with their operational needs. Our customers really appreciate this proactive approach.
How does the platform help with upcoming regulations and compliance audits?
Compliance is a significant burden for many facilities. Meeting Environment Agency reporting requirements usually involves extensive manual work, but the platform changes this by analyzing material directly at the infeed and key outputs. It automates the sampling process and provides the documented evidence needed for audits. By replacing manual reporting with real-time data, it allows plant managers to move away from administrative tasks and focus on making quick, operational decisions.
Where can recyclers and MRF operators see the new platform in operation?
Operators can test our platform at the TOMRA Recycling test centers in Mülheim-Kärlich, Germany and Charlotte in the United States.
Parshva introduces PolyPerception to Messe.TV at IFAT 2026
Messe.TV caught up with Parshva at IFAT 2026 in Munich for a short interview where he explains the power of waste analysis and its benefits for recycling plant operators.