.png)
Deeptech From Scratch #2: Dunia, the million material machine
Welcome #2 in our new series exploring the often-messy, sometimes-brilliant, and always-audacious journeys of founders tackling the world’s hardest problems. Each profile is released as a podcast episode alongside an accompanying article.
In this, our second ever episode, we get into it with Dunia. Founded by Ahmed Ismail, Alex Hammer and Marcus Tze-Kiat Ng, they've set out to close the ‘Sim2Real’ gap for catalysts R&D. To find out what this means, you can listen to the full interview with Ahmed here:
.jpeg)
When Ahmed speaks about the state of industrial materials discovery, he doesn’t reach for euphemisms. “It takes decades to find these materials” he says. “We are accelerating this to months”.
It sounds like a slogan until you see what Dunia is actually building: a fully autonomous factory-scale system designed to close the most fundamental data gap in catalysis and industrial chemistry.
This is not an incremental project. It is an attempt to compress the front end of global catalyst R&D (and a load of other materials) by 1-2 orders of magnitude. When you realise quite how crucial catalysts are and how slow their rate of progress is, this truly bites: all this and more we cover in our podcast and below in the article.
The collapse in materials science
For years, synthetic chemistry and materials engineering have quietly accumulated a structural failure: lack of standardised, machine-learning-ready data at scale.
In biotech, AlphaFold stood on a foundation of hundreds of millions of high-quality, standardised protein structures. In catalysis? Ahmed puts the true number plainly: “The largest industrial database you have right now is something like 500 to 900 data points” for CO₂ conversion.
To make things worse, the field suffers from an almost surreal lack of reproducibility. As he notes, “The chance of you yourself running your own experiments twice and reproducing your own results... is less than 50% these days”.
Materials research is, in effect, running on a coin toss. AI cannot learn from noise, and the industry cannot afford decades-long discovery cycles for catalysts that underpin $34 trillion of global industrial processes.
This is the vacuum Dunia is filling.
Why autonomous labs haven’t delivered
Legacy players have built automation. No one has built true autonomy.
Most “robotic labs” amount to high-throughput testing rigs: researchers make samples manually then load plates or reactors for automated characterisation. Dunia team realised this misses the point. Without control over synthesis, the data is still incoherent.
“If you try to use an AI today to look at the data from literature, it’ll never learn anything” Ahmed says. “It was made in very different ways”.
Dunia’s answer is a clean-room philosophy applied to chemistry: a closed-loop robotic environment that standardises every step from prediction to synthesis to testing, building high-fidelity datasets at industrial volumes.
Iris, Dunia’s third-generation platform, is the result. Iris already runs 56 complete make–test iterations per day, end-to-end. “Fully autonomous. No human intervention.” That number alone puts Dunia among the most advanced autonomous chemistry platforms globally, equivalent to 10% of the world's catalysis research workforce - in a single room. But it is nowhere near enough.
Dunia’s Giga Lab
.jpeg)
To train physically-grounded models that can generalise across complex catalytic systems, Dunia estimates it needs hundreds of thousands of experiments, not hundreds. That requires a qualitative shift.
Hence the Giga Lab: a 5,000-m² autonomous facility capable of 500 to 1,000 material iterations per day. Dunia plans to generate one million high-quality data points per year, far surpassing anything available in the field. “No one has that level of imagination yet. No one even knows how to do that.” Ahmed remarks. “It’s a factory that can think.”
The implication is direct: the first million-material dataset becomes the foundation for the first generalisable catalyst-design AI. Whoever builds it sets the pace for the next era of industrial chemistry.
System success to date
Platform success is dictated by market demand and validation of results. Dunia now has both. “We actually have two patented materials from Iris.” One is a hydrogen catalyst that surpasses what the even the core conclusions of Ahmed’s own PhD predicted was possible. The other is a conductive coating now under discussion with a major Japanese manufacturer.
On the commercial side, demand has already outstripped capacity: “We have signed tens of millions of commercial agreements... some binding for the next three years.” Eight parallel projects is Iris’s current operational ceiling, hence the urgency to build the Giga Lab.
How this impacts the climate
Every climate model hides a missing variable: breakthroughs in materials are assumed, yet absent. Electrolysers, CO₂ conversion reactors, green ammonia plants: all depend on catalysts with lower activation energies and higher selectivity.
“Hydrogen would become more green, hydrogen would become more comparative cost wise to grey hydrogen” if the right catalysts existed, Ahmed explains. “The problem is we don’t have these materials right now”.
The bottleneck is no longer engineering. It is discovery. And discovery, today, is too slow to meet the climate needs.
The origin
.jpeg)
Dunia did not begin as a spinout or a research artefact. It emerged from 16 years of industrial pain points.
“When you have a problem, we are talking about hundreds of millions of dollars a day of loss” in oil and gas operations, Ahmed recalls. And the best available response? “Sponsor a PhD for four or five years” with only a chance of relevant output.
The project developed out of the initial scoping completed by Gael Gobaille-Shaw (pictured above at Dunia's Berlin HQ), founder of Mission Zero Technologies (abundant CO2 from the atmosphere) and Supercritical Solutions (abundant green H2 in an industrially optimised format), who realised that the catalysis gap would hold back his overall vision of "Closing the Carbon Loop", creating hydrocarbon materials sustainably from sunlight, wind, water and air.
The Dunia team converged from Egypt, Germany and Malaysia when Adam Tomassi-Russell, a Senior Director at DSV, encountered Alex and Marcus at an COP26 event in Glasgow and brokered the intro to Ahmed, who was anyway exploring the space as a Founder in Residence at DSV.
They speak the same scientific language but bring different cognitive cultures: Alex was focused on AI for materials, Marcus on robotics for catalysts and Ahmed on industrial applications of catalysts. These turn out to be the 3 crucial underpinning components that distinguish Dunia. The hardest part was not interdisciplinary integration but cross-cultural one, coming from wildly different national backgrounds: “That was harder than the disciplinary differences” Ahmed notes. But once aligned, “it went extremely fast” .
Zero to one
Dunia’s cold-start problem was severe: they asked investors to believe in a robot that could build a catalyst that could validate the robot.
“You need to have a lot of visionaries, we were lucky to have Deep Science Ventures there… you need people who can imagine what might happen in five, ten years” Ahmed says. Recursion’s Chris Gibson was one of the earliest to recognise the parallel to his own industry transformation and became an investor and advisor.
Even before Chris Gibson was enlisted, through the year they spent with DSV before incorporation, the team were able to access early funding and acceleration from the Net Zero Technology Centre and from Anglo American Platinum (now Valterra Platinum), allowing them to break the chicken and egg loop typical of the start of something extremely ambitious.
What’s next
Dunia is already expanding into batteries, thin films and conductive coatings. It is scouting locations for Giga Labs across Europe, Asia and the US. “We are trying to map where the next Giga Lab will be. The first one is gonna be in Germany... we had a lot of offers already from different governments all over the world”.
The invitation is explicit: if you want a million-material machine in your region, now is the moment.
Dunia goes beyond incremental optimisation. It is instead compressing decades of material discovery into a continuous autonomous cycle, producing breakthroughs at industrial cadence. If successful, it resets the trajectory not just for catalysis, but for the economics of the energy transition itself. “A factory that can think” is not a metaphor. It is the new frontier of climate technology.
You can learn more about Dunia on their website and LinkedIn.