Will Purvis is the co-founder and CEO of Kaio Labs, a company using automation and machine learning to develop modular reactors that convert waste emissions into chemicals.
What led you to found Kaio Labs?
My background is in chemistry. I represented the U.K. at the International Chemistry Tournament in high school and studied chemistry at the University of St. Andrews. During one of the lockdowns I taught myself to code and started exploring the intersection of computer science and chemistry.
This led me to machine learning research at the University of Toronto, where I worked on what is called a “self-driving lab.” It’s not a lab on wheels, but self-driving in the sense that automation systems ran experiments and generated real-world data that fed into models to iteratively optimize the process. At the same time, ChatGPT came out, making me realize how impactful AI could be. To learn more, I went on to study machine learning at Cambridge and further develop my technical skills.
While at Cambridge, I started exploring how AI could be applied in the chemicals industry. Through those conversations, I realized that most chemical production still relies on fossil fuels and century-old infrastructure that’s struggling to decarbonize. Electrolyzers seemed like a promising alternative since they can turn waste emissions into the same chemicals, but they’re still far too expensive to compete. That’s when I realized I could use the same “self-driving lab” concept to optimize the material inside these electrolyzers, making them far more efficient and economical.
A weekly dispatch featuring exclusive interviews with deep tech founders & a roundup of the most important deep tech news.
How does Kaio Lab’s approach differ from traditional electrolyzer development?
Electrolyzers are devices that use electricity to drive chemical reactions, meaning that you don’t need extremely high temperatures or pressures to run a reaction. Traditional development in this field, whether for hydrogen production or other electrochemical processes often center on a single breakthrough within the electrolyzer such as the catalyst inside of it, the reactor itself, or system design.
Kaio labs takes a different approach. Instead of optimizing one component at a time, we use our automated lab and machine learning algorithms to co-optimize the entire electrolyzer stack across many variables simultaneously. This essentially allows us to squeeze as much performance as possible out of the entire device, making the conversion as economical as we can while also tailoring systems to different products and customer requirements.
Every experiment we run feeds back into our models, creating a growing data advantage. Our predictive models work like ChatGPT in that they learn patterns from large datasets to suggest high-performing catalysts or operating conditions. Our optimization algorithms then act as the decision-making layer, finding the best combination of parameters to achieve maximum performance and efficiency.
What makes electrocatalysis particularly suited to your high-throughput approach?
Catalysts are materials that decrease the energy required for reactions, and pretty much every chemical reaction is guided by a catalyst. Electrocatalysts specifically use electricity to guide the conversion process. The really neat thing is that electrocatalysis happens on a surface area, so it scales by surface area, not volume.
This means you can run experiments at a small scale in the lab, and when you scale up, it’s not a case of taking powders and making pellets and beads. It’s taking that surface area, making it bigger and stacking it. This is much more amenable to scale R&D results from high-throughput and automation systems towards commercialization.
Are you successfully cracking the economics of CO2 utilization?
The market need is clear. There’s significant investment in carbon capture, but ultimately all that captured carbon ends up stored in the ground with no tangible return on investment precisely because carbon utilization has been economically unviable. The literature and patent landscape show catalysts and reactor designs that theoretically make economic sense when scaled, but there’s been a reproducibility crisis where results either can’t be reproduced or fail to scale.
That’s why automation is so critical. It’s not just for generating data and increasing throughput, but for improving reproducibility and ensuring we can scale better. Over the past three months, we ran our first campaign targeting syngas, a mixture of hydrogen and carbon monoxide used to make fuels.
We’ve taken those learnings and started looking at other high-value chemicals like ethylene and formic acid, which are the backbone of chemicals we use daily. We established our lab in January and achieved state-of-the-art literature performance by August, essentially decreasing R&D timelines from multiyear campaigns to multi-month campaigns.
How do you position yourselves in the market?
We’ve seen numerous plant closures this year in the chemical space due to rising operational costs. Margins on chemicals are slim, and when you add carbon taxes and consumer pressure to decarbonize, it results in plant closures. We’re not competing with chemical companies; we’re providing them with a solution: They have all this waste CO₂, and we can run it through our optimized reactors so they can increase production capacity without paying carbon taxes and satisfying consumer demand for sustainable chemicals.
On the machine learning side, we want to deploy this technology at scale. The platform is our in-house tool to get there, but it’s not the company itself. It’s our approach and differentiator to get to market faster with highly optimized electrolyzer systems.
If Kaio Labs succeeds, how does the world look different?
The vision is to unlock decentralized chemical production: just like wind and solar have decentralized energy production, we aim to do the same with chemicals. There’s been a lot of talk about supply chains with tariffs and the Suez Canal affecting these huge chemical complexes centralized in the Gulf Coast and Rhine region. Our technology enables modular chemical production wherever emissions are present, creating a distributed network that minimizes reliance on complex supply chains or exposure to geopolitical risks.
Equally important is the sustainability impact. Our chemicals are net-zero and made from waste carbon dioxide, moving the industry toward electrified production systems. Just as the energy and transportation sectors are shifting to EVs and solar, the chemicals industry can transition away from fossil resources. While costs have historically made this challenging, our platform is designed to make this transition possible.
How do you define deep tech?
When I was in kindergarten, what we now think of as software was probably deep tech then. The bar keeps moving higher. To me, deep tech is really the frontier of science and engineering, the ability to take different, maybe seemingly unrelated concepts and put them together to create a technology that has potential for significant impact on the world. It’s the frontier of science and engineering that, if we can achieve it at scale and it succeeds, will have a major impact on the world.