AsiaTechDaily – Asia's Leading Tech and Startup Media Platform
“A drunk man staggers out of a bar and starts searching for something under a streetlight.
His friends approach him and ask, ‘What are you looking for?’
He replies, ‘My car keys.’
They ask, ‘Where did you lose them?’
He points to a dark corner and says, ‘Over there.’
Puzzled, they ask, ‘Then why are you looking here?’
The drunk squints and confidently responds, ‘Because this is where the light is!'”
AI is Set to Transform Human-Biased Experiments
“This joke perfectly illustrates a common problem,” says Hermann Tribukait, co-founder and CEO of AI startup Atinary Technologies, as he begins our interview with this vivid analogy. “When searching for new materials, whether catalysts or new drugs, researchers are often guided by their biases or limited knowledge, only looking in familiar places. They are often guessing, applying trial and error and don’t realize they’re searching in the wrong spot and missing other possibilities.”
Chemistry is fundamental to modern life and various industries. Everything around us involves a chemical reaction. All chemical reactions require energy. Catalysts are crucial chemical compounds making reactions faster with less energy, and allow for new reactions and materials.
According to the Nobel Foundation, 90% of industrial-scale chemical reactions use catalysis, and 35% of global GDP is based on catalysis. Without catalysts, everyday products like medicines, fertilizers, plastics, petroleum, fragrances, and food wouldn’t be possible.
Developing next-generation catalysts drives modern progress. Without new catalysts, industry and technology would stagnate.
However, developing new materials and catalysts is time consuming and expensive. Take new drug development, for example. From drug discovery and lab research to clinical trials, FDA approval, and finally hitting the market, the whole process can take more than a decade. In terms of cost, the investment required to develop a new drug can exceed US$2 billion. Even more discouraging is the low success rates between 5% and 10%.
“It surprises me that most R&D is still human-driven,” Tribukait explains. Many corporate R&D departments and university labs haven’t adopted AI and automation, still relying on tedious, repetitive tasks performed manually, using Excel spreadsheets, or even worse, pen and paper records.
Developing new materials involves screening extremely complex chemical spaces. The more variables there are, the larger the number of combinations. Finding the optimal solution among millions or even billions of possible combinations is like finding a needle in a haystack. No wonder the process is time consuming, costly, and inefficient.
This is the pain point Atinary aims to solve. The name “Atinary” comes from the verb in Spanish “atinar”, which stands for “hitting the target.” “We developed a no-code AI platform that significantly accelerates R&D, optimization and discovery of new molecules and materials, making the process more efficient and effective, helping scientists hit their hard-to-find targets,” says Tribukait.
A startup sparked by the collaboration between chemists and computer scientists received angel investment in its very first month.
Atinary was founded by a multidisciplinary team. Hermann Tribukait, a Harvard PhD in economics and an entrepreneur, developed international public-private partnerships to promote innovation and led a global initiative in accelerated materials discovery prior to launching Atinary. Tribukait built a global network in academia, governments and industry in AI-driven R&D. Cofounder Loïc Roch, a quantum chemistry PhD from the University of Zurich and Tianjin University in China, conducted postdoctoral research in chemistry and AI at Harvard and in Toronto. Roch has published over 50 papers in top international journals, establishing himself as a thought leader in AI for chemistry.
In 2017, the two met at an international conference on accelerated R&D organized by Tribukait and immediately clicked, sharing the belief that AI would revolutionize materials discovery and innovation. In 2019, Tribukait and Roch co-founded Atinary and secured their first angel investment within the first month.
Roch developed a cloud-based no-code AI platform. Scientists don’t need programming skills; after just two hours of onboarding, they can deploy machine learning models with the click of a button in their existing workflows to speed up optimizations and their R&D overall.
No-Code/Low-Code platforms have become increasingly popular in the past two years. In the case of Atinary’s no-code AI platform SDLabs, even users with no coding and no AI knowledge can use it easily. According to a report by global technology research company ISG Information Services Group, the global market for No-Code/Low-Code development platforms is nearing $15 billion and is expected to quadruple in the next five years.
“Life is short, and we have limited time to learn science and chemistry. To develop good products, researchers might also need to learn computer science and AI,” says Tribukait. However, with SDLabs, scientists don’t need to code. They just need an internet connection and defined optimization problems to run complex experiments that traditional methods can’t handle, finding the best solutions in a fraction of the time and cost.
“The human brain is not suited to screen the massive and complex chemical spaces in experiments. We can do much better if we leverage the power of computers and algorithms, which is why no-code platforms are crucial to accelerate scientific innovation,” Tribukait concludes.
What is Atinary’s “Self-Driving Labs Platform”? Imagine You’re Baking a Cake …
Atinary has developed an AI platform called “Self-Driving Labs Platform” or SDLabs, aiming to guide research just like autonomous driving guides a car. The difference with self-driving cars is that the human remains in the loop and at the center of the experimentation. The SDLabs platform augments researchers to solve complex experiments much faster.
Atinary’s founders, Dr. Hermann Tribukait and Dr. Loic Roch, defined Self-driving labs as follows: “Self-driving labs are innovative experimental platforms that fully integrate artificial intelligence, robotics and digital technologies.”
Self-driving labs are revolutionizing how science and R&D are done. We can accelerate R&D pipelines and time-to-market of new breakthrough products. Addressing the challenges in chemistry, advanced materials, and other fields including pharmaceuticals, these technologies can help the transition to a circular economy.
Imagine you want to bake a cake, you know the ingredients that are available in your kitchen but you don’t know the best recipe or formula. Which recipe (ingredients and amounts) is the best? What temperature should you use? What ingredients should go in first? You have no idea.
So, you decide to wing it: “Let’s just start! Maybe I’ll accidentally create a delicious, never-before-tasted cake!”
Sure, you might get lucky, but the number of potential combinations is huge, your budget is finite, and the clock is ticking. Guesswork isn’t the best strategy.
Then, a smart assistant appears, offering suggestions you never thought of. You try it and taste it. Your assistant learns from this first attempt and provides a new recipe to make another cake.
Wow, it’s much tastier than the first! But the second cake still isn’t perfect. The assistant tweaks the recipe again based on the second cake, and you keep improving.
This iterative process continues. With each adjustment, the assistant gets smarter, saving you guesswork. Together, you create the perfect cake after a few iterations.
“Atinary’s SDLabs works similarly,” Tribukait elaborates. SDLabs uses sequential learning strategies,” a machine learning framework combining Bayesian optimization and active learning, adaptable to various algorithms. In less than two hours of onboarding, scientists can deploy Atinary’s machine learning technology in their existing workflows.
SDLabs guides the experiment, solving multi-parameter and multi-objective optimization problems, analyzing data, and making AI-driven experimental decisions. Crucially, SDLabs explores unconventional and uncharted research directions. It not only boosts productivity but also accelerates the discovery of new materials and molecules, significantly reducing the number of experiments required to reach global optima and costs. Ultimately, it leads to new knowledge and enables new technologies, optimizing experimental processes and speeding up product launches.
The team behind Atinary’s innovative platform is a diverse group of highly skilled professionals. Atinary’s platform was created from scratch by a multidisciplinary team of 16 experts, spanning AI & ML engineering, chemistry, biochemistry, molecular biology, physics, software engineering, DevOps, and data science. The team’s cross-disciplinary expertise has been key to developing a platform with the potential to transform R&D across industries. Atinary’s innovative approach is bolstered by its distinguished Business and Scientific Advisory Boards, featuring leading figures in supercomputing, AI & ML, computer science, chemistry, entrepreneurship, and industry leadership. With this strong foundation, Atinary is set to revolutionize the application of AI in various scientific fields.
SDLabs is offered using a standard software-as-a-service (SaaS) B2B business model, initially targeting industries like pharmaceuticals, biotechnology, and chemicals. Its users have already run 100,000 experiments on SDLabs. Atinary has deployed its technology with notable organizations including Takeda, Japan’s largest pharmaceutical company, DSM-Firmenich, MIT, ETH Zurich and IBM Research, among many others. An MIT Soft Materials Lab professor noted that tasks that would have taken two years were completed in a week with SDLabs.
Speeding Up R&D: SDLabs as the Scientist’s Time Machine
SDLabs is like the AI in self-driving cars that helps with navigation, observation, and decision-making. However, Tribukait emphasizes, “We’re not replacing the driver; scientists still hold the steering wheel. The algorithms make predictions and drive the process, they learn from the experimental data produced by the scientists on the bench, or by robots, or by computers in the case of simulations, and make the data-driven decisions on the next iterations to run. SDLabs also builds the data libraries and offers data analytics tools that shed light on the AI decisions and on the results.”
Beyond accelerating R&D with AI, Tribukait and Roch are driven by their love for nature. Tribukait shares emotionally, “Both Roch and I are nature enthusiasts—I love the sea, and he loves the mountains.” We can accelerate the solutions to critical challenges, including pollution and climate change, by developing new high-performance materials and catalysts that reduce the use of harmful chemicals and waste. Atinary hopes SDLabs can be a “time machine” that not only speeds up R&D but also lessens the environmental impact.
“Going back to the joke about searching for car keys, with Atinary’s no-code AI platform, scientists won’t have to fumble in the dark. AI will guide the research, screening the chemical space efficiently and uncovering possibilities they never imagined,” Tribukait says. The vision is to change how R&D is done and democratize the access to AI-driven experimentation. “We believe that AI will revolutionize R&D across all industries.”
This article is part of a partnership with Cherubic Ventures. Founded in 2014, they are an early-stage venture capital firm that’s active in both the US and Asia, with a total AUM of 400 million USD. Focusing on seed stage investments, Cherubic aims to be the first institutional investor of the next iconic company and back founders who dare to dream big and change the world.