CircularNet: How Recykal built Asia’s largest circular economy marketplace using Google AI

September 2023

Featured technology

CircularNet, Google Cloud, AI models, Maps

Who we’re helping

Local communities, governments, businesses

Our role

CircularNet is being used to help save valuable recyclables from landfills and curb plastic pollution

Recykal_Google Sustainability Image_900x506 pix_02.png

India has one of the world’s fastest growing economies, yet still lacks a robust recycling infrastructure. Even the smallest recycling facilities globally process tens of thousands of pounds of material every day. In Delhi, around 30 million metric tonnes of landfill waste sits in the city’s notorious garbage towers. Not only does this add to the city’s air and land pollution, local residents are also faced with polluted water supplies. Local landfills are caught in a vicious cycle, unable to keep up with sorting incoming waste for recycling and composting, on top of the excess left to pile up over the years.

“India produces around 62 million metric tonnes of waste a year,” says Chetan Baregar, associate director of marketing at Recykal. “But its waste management ecosystem is still in its nascent stages.”

Based in the southern city of Hyderabad, Recykal’s 500-strong team is rethinking India’s recycling ecosystem. Powered by Google AI, the startup has created the fastest-growing waste commerce marketplace in Asia.

Through CircularNet, Google’s open-source machine learning model for waste management, Recykal says it has achieved more than 90% accurate detection of metals, bottles, tubes, cutlery, PETE, HDPE and LDPE plastics. In turn, their measured production quality has improved by 60% through the use of Google AI.

Baregar says recyclers across India are receiving pure dry waste without contamination—which increases the financial value by roughly six-times. Plastic, paper, metal, e-waste and batteries can all be sold on Recykal’s B2B marketplace, now operating in more than 30 Indian states and union territories. But the biggest benefit is societal: Recykal now says it diverts 50,000 metric tonnes of waste from landfill every month.

“If material isn’t going to be recycled, then it’s going to eventually end up as land, water or air pollution,” says Baregar. “Using the CircularNet AI technology, we’re helping brands reach their sustainability goals, providing governments with actionable insights, and sourcing businesses better and cheaper materials. The technology is helping to benefit India as a whole.”

Unlocking value in waste management

Waste is a big business. The global waste management market is valued at nearly $1.3 trillion, with the world population generating an estimated 2 billion metric tonnes of rubbish every year. The bulk of the industry is made up by recycling: materials are collected from households, processed at Material Recovery Facilities (MRFs), then sold to vendors.

As it stands, only a fraction of global resources ends up recycled. In India, separating valuable recyclables like e-waste, metals and plastics from non-recyclables is largely done by hand. This is due to recycling contamination, which happens when disposed items are put in the wrong recycling bin, or food residue is not cleaned off. As a result, recycling rates can be inefficient as workers must hand-pick which material can be recycled.

This doesn’t just impact profit margins for waste management operators—it harms the environment, too. Recyclable material can end up in waste landfills by mistake, and contaminated items can disrupt the recycling process.

Incentivizing better recycling habits

This is where Google’s AI technology, CircularNet, comes in. The open-source machine learning model helps to identify and reduce contamination in recycling streams, for increased landfill diversion. Images are captured at different points inside of MRFs, generating daily metrics and enabling a data-driven approach to recycling. In turn, the technology is helping Recykal to deliver actionable insights to key stakeholders, from processors to governments and major brands as India’s leading Producer Responsibility Organization.

“CircularNet was the perfect fit for us, so we didn’t have to build a solution from scratch,” says Sanket Waghmare, product owner of SmartSkan at Recykal.” It provided us with the technology to detect categories of waste with precision, allowing us to focus on building data models—and immediately go-to-market. We could just build on top of it, and get to work on India’s environmental and public health goals right away.”

Supporting the waste management process from end-to-end, CircularNet is used to identify materials with precision, measure contamination, and enable Recykal’s partners to circulate more and better quality recyclables back into the supply chain.

“We’ve become a kind of thought leader when it comes to recycling in India,” says Chetan Baregar, associate director of marketing at Recykal. “Any product made from plastic is sold with a deposit on top of 5-10₹ [approximately 5-10 pence]. It’s promoting packaging as a form of currency, rather than something that should be wasted.”

Accelerating data breakthroughs

With Recykal’s Smart Skan solution, almost 13,000 households have been provided recycling bags with unique identifier codes. Weekly household collections are transported to Smart Centers and MRFs, and using Google AI, cameras scan collected materials emptied onto conveyor belts.

Baregar says these insights are crucial in supporting India’s waste management ecosystem. “Before CircularNet, authorities lacked information on what kind of waste was being generated, how recycling habits differed across India, and how to plan and budget accordingly.”

“Now, municipalities are able to identify where they need to focus greater consumer education, or incentivize recycling behaviors. Better-value materials are provided to buyers, and AI means we’re able to scale our capacity, which helps to increase profitability for waste management centers.”

If you are interested in collaborating with us on this journey, please reach out to