A team is at work in West Germany to create a circular economy hotspot akin to California’s Silicon Valley. Startups with innovative ideas, established companies the likes of Bayer, and other stakeholders wanting to advance circularity physically convene in this booming region to both gain and give insight.

Part One of this series looks at the inner workings of the project, dubbed as the Circular Valley Initiative. Part 2 is the story of a participating startup; it tells where this French company was on its journey when it got involved in the initiative, and where it is now. 

French startup Lixo has tailored artificial intelligence (AI) technology for the waste management world, providing real-time intel on materials collected; spotting sorting errors; and detecting new objects making their way into the stream, then adding them to its software so they can be spotted and plucked out.

Marjorie Darcet, co-founder and CEO of Lixo, further describes the smart technology; who some of the young company’s not-so-young or small customers are; and a likely roll out into North America soon.

Darcet also explains what inspired the early brainstorming to develop the concept— tease: it has to do with a prototype robot walking the banks of a canal in Paris that fast caught attention on social media and appeared on local TV.

Waste360: What are the components of your system and how do they work?

Darcet: Lixo integrates three tech bricks:

1. sensors that capture images, placed in trucks, on conveyor belts, or unloading halls;

2. image analysis software that detects each object and characterizes it;

3. data visualization tools (dashboards and APIs) that display waste analysis in understandable and leverageable ways.

Waste360: Exactly what information does your technology provide and what can your clients do with it?

Darcet: They can get precise, real-time data on the waste being collected [within collection trucks, materials recovery facilities (MRFs) or treatment facilities].

They can act on billing (charge more for clients not respecting sorting rules or providing below-quality material). Municipalities can better interact with inhabitants. And they can adapt their operations to the quality of waste flows.

Waste360: How does your analysis help improve the quality of waste collected?

Darcet: There are 2 ways Lixo helps haulers:

● in real time: provide alerts and re-route trucks when contaminants or hazardous items are collected (to prevent sending them to treatment facilities where they can harm machines and workers);

● on the long run: communicate with clients and inhabitants that do not follow sorting rules, change the billing of clients with higher rates of contamination, and therefore reduce contamination.

One example is the work we did with a city in Western France.

With our technology they identified the most frequent and contaminating sorting errors in the recyclables stream; identified the location of most recurrent sorting mistakes; built targeted communication campaigns; and increased by more than five percent the sorting rates quality (equaling more than 1,000 metric tons diverted from incineration that got recycled, and 2,000 metric tons of residue not sent to MRFs).

Waste360: What are the biggest problems you help solve?

Darcet: The biggest problem we solve is the lack of data in the waste management industry. With the increasing role played by regulation, on both sides of the Atlantic, there is a major need for an increase in waste management efficiency. But one can only improve what one can measure.

Providing data on waste quality along the value chain is key to: informing stakeholders [including emerging Extended Producer Responsibility (EPR) systems]; improving operations; and making price match quality.

Waste360: How do you adapt the technology to keep up with evolving streams, changing markets, and changes in materials and design of packaging?

Darcet: Contrary to Near-infrared (NIR) technologies, which can struggle with working on mixed materials, computer vision (AI algorithms) can adapt to new flows continuously. Because images are sent to our systems, we can detect new types of objects and packaging, integrate them in our software, and update our sensors remotely.

Waste360: Describe your journey from your start till today

Darcet: We started working on MRF and recycling facilities (analyzing waste over conveyor belts). Our objective was to smoothen the relation between MRFs selling sorted material and recyclers buying that material. But we soon realized that these facilities struggled to produce quality material because they did not have control over what was collected and brought to them.

We launched our first collection pilots in 2021 and now work with all major players in France and operate in four European countries, with a total of over 30 clients (including Suez and Veolia). In 2023, we’ll expand to three more European countries and most likely sign our first contracts in North America.

Fun fact: Lixo originated as a side project when our CTO, Olivier, was a data engineer at a French tech firm. In his free time, he built a robot that walked around with a small camera on its “head,” detecting cigarette butts thrown on the ground and vacuuming them. The robot got attention on social media and even appeared in local TV news. It led Olivier to dig into what artificial intelligence could do for waste management at an industrial scale.

Waste360: Tell us about plans to expand in North America

Darcet: Over the past few months, we were contacted by several North American players, including some very large companies. We expect our first projects to launch by the end of the first semester of 2023.

We believe our experience with leading EPR companies across Europe (Citeo in France, PontoVerde in Portugal, and Ecoembes in Spain) can benefit players involved in emerging EPR systems in North America.

Waste360: Why did you decide to develop AI specifically for waste management pros?

Darcet: Because very few people were doing it. So far, in the waste management industry, AI has mainly been used for robotics. But it can do much more, like help make commercial decisions, adapt prices in real time, etc.

AI also can adapt to fast-changing environments and be replicated. We started by working on recyclables (paper, plastics, metals) but now also work on residue and biowaste, which is a booming segment.

Waste360: What challenges unique to waste did you consider?

Darcet: We deal with very complex, overlapping objects, whose shapes and colors have been dramatically altered by transport, handling, and piling. Our challenge is to make sense of all that and deliver actionable insights.

Waste360: How much financial backing did you begin with and who were your backers?

Darcet: We raised a 3.5m€ ($3.8) seed round in early 2022. Prior to that, we bootstrapped to launch pilot projects and received funding from France’s public investment bank to support our R&D.

We were lucky enough to be backed by three complementary investors: Raise Ventures, very active in the software and impact field; Demeter, with extensive experience in the energy and waste management sector; and Amaury Bierent, with over 30 years’ experience in the waste management and environmental sector.

Waste360: What would surprise you in a good way if it were to happen?

Darcet: We were happily surprised to spark the interest of North American players so early on. We were initially planning to wait another year before expanding, but I think the uniqueness of our solution and the fast evolutions of the North American market accelerated things.

So, if we were to reach our revised goal of expanding to the U.S. by the end of the first semester of 2023, that would indeed surprise me in a good way, and reward the hard work we’ve put into developing Lixo.

Waste360:  Where do you see AI going in the future in the world of waste?

AI in waste, as in other industries, will have to prove its ability to adapt to concrete situations and be integrated with existing processes. One of the common traps is to use AI because it is appealing and available. But AI can quickly become more of a gadget than an actual tool.

So for the future of AI in waste, I see collaboration and combination: AI will have to be integrated with other technologies (NIR, etc.) and lower-tech tools.