AMP Robotics has come a long way since its 2014 launch. The company that first brought artificial intelligence (AI) and machine learning to a few materials recovery facilities (MRFs) now boasts a platform that recognizes over 75 billion objects, which AI-guided robots pluck from the stream at astonishing speed.
AMP has been lauded for its “world-changing” ideas by Forbes and called out as a “circular economy tech disruptor” by the World Economic Forum, among a showering of special recognitions. The company has lassoed in $180m in funding and an expanding list of contracts with heavy hitters in the recycling world.
In this Q&A, Matanya Horowitz, the company’s founder and CEO, takes us back to the company’s beginnings, up to today, pausing on a few milestones and key partners, including some industry names we all know. And he tells us how he sees AI evolving to meet growing demands placed on recyclers.
Waste360: Tell us about AMP’s beginnings.
Horowitz: While studying for my PhD, I saw some of the major results in the subject now known as deep learning. A series of algorithmic breakthroughs led to machines learning how to see, roughly as well as a human. After graduating, I wanted to find places where this technology could be useful, and I found that in the world of recycling.
When I visited recycling facilities, operators told me about issues with the quality of the materials being separated out and high rates of staff turnover. The convergence of machine learning and robotics offered compelling opportunities to automate what had historically been tasks that were labor intensive, high cost, inconsistent, and limiting, while unlocking significantly more value from these complex, heterogeneous material streams.
Waste360: What are some of your greatest milestones?
Horowitz: The milestones have been many since the company’s founding in 2014 and we began selling to early adopters. In 2020, we signed our largest deal to date with Waste Connections, and in 2021 we expanded to Europe and piloted next-generation secondary facilities.
We’ve grown our product portfolio to include solutions for material characterization, film removal and recovery, and compact spaces in MRFs—all enabled by AI.
Now, we’ve introduced a fully automated, facility-scale sortation solution that integrates all of AMP’s AI-powered sortation technology into custom or turnkey facilities to deliver maximum resource recovery with minimal labor and cost.
Waste360: What do your flagship products do? And what distinguishes each?
Horowitz: AMP Cortex is our high-speed, AI-guided robotic sorting system. Robots perform tasks of sorting, picking, and placing material to achieve up to 99% accuracy and 80-120 picks per minute. AMP Cortex-C is a compact version of this sorting system that adapts to space constraints and brings our AI expertise and robot technology to more locations.
AMP Vision is a modular computer vision system designed to drop into key stages of recycling operations to better understand material flow.
AMP Clarity is our portal for recycling data and insights and robot optimization where users can monitor real-time material characterization and performance measurement throughout a facility.
AMP Vortex is an AI-powered automation solution designed to tackle film contamination and improve recovery of film and flexible packaging.
AMP Microjet is our most compact, easiest-to-install sortation solution; it employs air nozzles that fire from the side of a conveyor belt for high-efficiency sortation on lower-volume lines.
Waste360: How do individual operators decide which solution would be best for them?
Horowitz: AI provides a universal sensing platform for recovering value from waste streams, but the best physical mechanism may depend on the material type. The robot is a Swiss Army Knife that can handle most materials, whereas Vortex is an example of recycling-specific automation that’s designed for thin film. Combined, our technology suite can tackle the majority of non-automated sorting stations in a MRF, and without a significant retrofit to existing infrastructure. Moreover, we’ve integrated these technologies into a comprehensive facility solution that’s helping lower processing costs, enable more profitable sorting, and ultimately, expanding recycling access and infrastructure.
Waste360: How much money have you raised and how are you investing it?
Horowitz: We’ve raised more than $180 million to date. We have a number of larger opportunities in the U.S., Europe, and globally such as large, fleet-wide deployments of robots and deployment of fully automated sorting facilities. The capital helps us build the technologies and team to support these opportunities.
Congruent Ventures and Wellington Management led our Series C round; we also have support of investors including Blue Earth Capital, Microsoft Climate Innovation Fund, Sidewalk Infrastructure Partners (SIP), Tao Capital Partners, XN, Sequoia Capital, GV, Range Ventures, My Climate Journey, and Valor Equity Partners.
Waste360: Who are some of your partners and how do you work together?
Horowitz: Our technology is deployed in nearly 90 facilities. Our customers range from large, national players like Waste Connections – who we’ve partnered with to deploy more than 50 systems – to regional, independent operators to municipalities like Boulder County in Colorado and Emmet and Kent Counties in Michigan.
Waste360: AI has been around a long time in other industries. How did you make it work for recycling?
Horowitz: Although the broader field of AI has been around since the 1970s, the specific innovations we rely on in deep learning happened around 2013, which is why you saw such major advancements in recycling, autonomous driving, facial recognition, etc., around 2015.
Our technology perceives images of objects on conveyor belts within recycling facilities. Looking for specific colors, shapes, textures, logos, and more to recognize patterns correlated with material type, our AI learns to identify objects in the same way a human does.
However, recycling involves infinite variability in the kinds, shapes, and orientations of objects. Training a neural network to detect objects in the recycling stream is not easy, but it’s an entirely different challenge when you consider the physical deformations that these objects can undergo by the time they reach a recycling facility. They can be folded, torn, smashed, or partially obscured.
Waste360: How have you trained your system to do its job?
Horowitz: We’ve trained our AI platform using millions of images of containers and packaging types in different permutations.
AI accuracy and performance is built on training a platform with as many real-world images as possible. And AMP now has the world’s largest data set of recyclable material images for use in machine learning. Our platform recognizes well over 75 billion objects on an annual basis—a number that continues to exponentially increase. AI allows for increased specificity in what's sorted, along with the ability to deliver more consistent chemistries. We’re expanding the set of materials that are readily recyclable, as well as paving the way for new recycling technologies that can accept these materials.
Waste360: How do you see AI fitting into policy around materials management?
Horowitz: AI systems can profoundly change the way MRFs recover materials—and revolutionize the way policy makers view the recycling industry. These systems offer adaptability to evolve along with the ever-changing recycling stream, giving MRFs flexibility in handling new packaging types as they’re introduced. Their continued exposure to new and recognized objects alike makes these systems incredibly powerful and adaptive.
With AI and computer vision, we’re able to identify and describe the objects that pass through MRFs and transform them, both desired commodities and undesired contaminants, into data. Digitizing the waste stream provides transparency about what is and isn’t getting recycled to create better solutions on the path toward a circular economy.
Waste360: Do you see AI helping to meet expectations of states with extended producer responsibility (EPR) programs?
Horowitz: As more states pass laws related to EPR for packaging, the producer responsibility organizations (PRO) as well as the regulators tasked with overseeing these programs will be faced with the challenge of how to collect, process, and make sense of the vast array of packaging types and brands for which manufacturers are responsible.
AI systems are already identifying and categorizing the material coming down the line; they provide a ready data repository for information PROs need to manage EPR programs, which place reporting requirements on stakeholders. AI equips these stakeholders with the same facts to hold discussions on what’s going well, and what needs improving. Our ability to upgrade existing infrastructure means that the industry can rapidly adapt to EPR schemes. AI can help us account for material and recycling markets not only as they exist today, but how they'll look in the future.