The global waste crisis is a staggering $1.6 trillion problem, with mountains of discarded materials ending up in landfills and oceans. But what if we could see this “waste” not as trash, but as a valuable resource? This is the mission of Ambarish Mitra, co-founder and CEO of Grey Parrot. After a successful journey in augmented reality with his previous company, Blippar, Ambarish pivoted to tackle a more tangible and pressing global issue. Grey Parrot uses sophisticated AI and computer vision to analyze and sort waste streams in real-time, bringing unprecedented intelligence to the recycling industry. In this conversation, Ambarish discusses the technological challenges of deploying AI in harsh industrial environments, the importance of building a cost-effective hardware and software solution, and how data is key to unlocking a truly circular economy where materials are recovered and reused, not discarded. It’s a fascinating look at the intersection of deep tech and environmental sustainability.
→ Enjoy this conversation with Ambarish Mitra, on Spotify, Apple, or YouTube.
→ Subscribe to ournewsletter and never miss an update.
Nataraj: What is Grey Parrot, and how did the idea start?
Ambarish Mitra: Grey Parrot is a waste intelligence platform that uses computer vision-based AI blended with material sciences to recognize large-scale waste flows. When people throw away rubbish, it ends up in material recovery facilities where it’s processed and sorted for recycling, landfill, or incineration. Grey Parrot uses analyzer boxes to recognize 100% of the waste flowing through these plants, helping to sort it more efficiently. It’s a large and complex problem because humans generate garbage at such a massive scale that it can’t be solved with just human or mechanical interaction alone. It requires a large amount of vision-based processing and was almost waiting for the AI era to kick in to address it. We saw a large, unaddressed opportunity. Plus, waste is a global crisis that impacts lives and the planet, so we decided to address this issue head-on.
Nataraj: Was the initial idea to do what you’re doing today, or was it different?
Ambarish Mitra: It was different. My co-founder and our initial team came from my previous company, Blippar, where our mission was to build the world’s first visual search engine. We built a large-scale vision model, but we realized our revenue model led to recognizing brands that often ended up in the bin. This got us thinking. Everyone has mapped the consumption world—Amazon, DoorDash, Instagram all know what you’re about to purchase. But after that $23 trillion of annual consumption ends up in the bin, there was almost no digitization. I call it the shadow economy. One reason waste remains waste is that no one is doing enough digitally to value and recover it. That’s why so much value is lost. So the idea came: why don’t we use our vision expertise to do something more impactful and circular? We call it waste, but we see it as paper, aluminum, and different types of plastic. We think of ourselves as a material asset recovery company rather than a waste company.
Nataraj: What is the actual product that you’re selling to companies in the recycling ecosystem?
Ambarish Mitra: Let me give you a brief intro to how waste works. Waste is thrown in bins, collected by trucks, and taken to Material Recovery Facilities (MRFs). It’s tipped out, piled onto conveyor belts, and goes through layers of mechanical processes. There are large leakages in that process, and a majority of that leakage ends up in landfill. Our goal is to reduce that leakage. We built hardware we call the analyzer. The job of the Grey Parrot analyzer is to analyze 100% of the waste flow in real time. These are rivers of waste on belts two meters wide, moving at three meters a second, processing up to 1,500 tons of waste per day.
When the camera recognizes 100% of the waste flow, it helps plant owners understand the unit economics of their business—what material comes through and what its financial value is. Secondly, it provides waste analytics to show if the plant is efficient or inefficient because every percentage difference is a revenue opportunity. The last thing is quality control—the purity of the materials. The more single-stream a material becomes, the more a buyer will pay for it. Finally, we’re integrating a brain into these mechanical machines, much like Waymo makes existing cars into self-driving cars. We are making these plants semi-automated by applying intelligence to existing mechanics, sending signals from one gate to another to ensure everything is sorted as purely as possible. The plant owner sees a dashboard where all this data is available, showing if the plant is working optimally.
Nataraj: What are the architectural and structural issues specific to this industry that you had to navigate? It sounds like you’re shipping hardware and software into environments that are not known for being tech-savvy.
Ambarish Mitra: That’s a great question. This is not a category where you can grow at any cost. It’s a cost-prohibitive industry where every cent matters. Unlike growth-oriented industries like e-commerce or advertising, you can’t have a variable cost architecture where revenue compensates for growth costs. Here, we have to recover more waste and create value from it. The tonnages are massive. So, we had to build an architecture where a lot happens locally on the machine. Our deep learning models sit locally so our costs don’t go up as we process hundreds of millions of images. We process images at the scale of social networks, but we’re processing trash, not people.
It also needs to be near real-time, because the system has to react within 30 milliseconds to trigger a robotic arm, an optical sorter, or stop the plant for hazardous materials. The system cannot rely solely on internet connectivity. We came up with an architecture that requires the internet periodically, but a lot of the processing is on the edge. A huge amount of the vision processing actually happens on the camera itself to normalize images, because lighting conditions in every plant are different. We built one platform that works in every plant. It was an interesting challenge to consider everything from image capture to model building to ensure it works with 99% efficiency, 24/7.
Nataraj: Can you talk a little bit about customer acquisition? How did you approach your first five to 10 customers and how do you scale now?
Ambarish Mitra: As an outsider, we had to learn the hard way. We came from a background of large-scale, vision-based compute, but we didn’t understand waste. So, in the first days, we did something smart: we built the first version of the product *with* the waste industry. We asked waste management companies what problems they were trying to solve, like counting for audit trails or quality control. We learned from them and released our first version by talking to seven or eight customers, giving them the intelligence for free for the first two years while we built our larger model.
We also didn’t build it in just one geography. We spread out across Europe, America, and South Korea to get diversity of data. Commercially, we started with a direct sales model, hiring people from the industry. Then we learned there’s a whole middle tier of specialized salespeople who are plant builders. They were already aggregating multiple technologies to build a plant, so it made sense to partner with them. In the last two years, we partnered with Bolograph, the world’s biggest plant builder, and Van Dyke Recycling Solutions in the US, America’s largest. We disintermediated our direct sales model through these strategic partnerships, which made us more cost-efficient and allowed us to scale effectively.
Nataraj: Which countries are doing the best when it comes to waste management?
Ambarish Mitra: Japan and Korea are very good. Germany is very good. The society is very conscious, and it’s designed to collect waste in many forms, not just from bins. Germany has a direct deposit scheme where people can return bottles for vouchers, for example. I would say there are four components to solving this. One is the manufacturer, who can take more responsibility through standardization, like how USB cables were standardized. Then you have the government’s role, which can enforce regulations. Then you have the waste management side, which can optimize and digitize with AI. And the last quadrant, which has a lot of power but often doesn’t use it, is the consumer making choices that are more circular in nature. Today, consumers are making some choices, governments are doing something, and a few brands are doing a few things in fragments, but a perfect storm hasn’t happened yet.
This conversation with Ambarish Mitra offers a compelling look at how advanced AI can be applied to solve one of the world’s most fundamental environmental problems. Grey Parrot’s innovative approach not only enhances the efficiency of recycling but also provides the critical data needed to build a sustainable, circular economy for future generations.
→ If you enjoyed this conversation with Ambarish Mitra, listen to the full episode here on Spotify, Apple, or YouTube.
→ Subscribe to ourNewsletter and never miss an update.
