Azulene turbocharges chemistry simulations with “better data, better models”
Using physics and machine learning, the company is accelerating computational chemical engineering.
By Ruhani Chhabra
Believe it or not, expert chemists are often working blind.
In most areas of engineering, scientists rely on virtual models to guide their designs. For example, before building a skyscraper or a bridge, engineers run computer models to test exactly how it will behave. But in chemistry, it’s different. Chemists still have to go into the lab and test things from scratch — often through a slow, trial-and-error process.
Azulene Labs is working to change that.
The biotech-and-chemistry startup is building machine learning models to predict how molecules behave, based on highly detailed physics data generated by the team. Unlike other approaches that use public or approximate datasets, Azulene starts with cleaner data derived from complex quantum mechanical equations, resulting in better predictions.
“We use very high-fidelity physics calculations to create datasets ourselves that are more accurate than anybody else’s,” says Nicolas Sawaya, founder and CEO.
Azulene’s models are already showing promising results in areas such as small-molecule drug development. Their early tests suggest the models can train much faster than traditional ones while being just as, if not more, accurate.
The idea for the company came from Sawaya’s background in quantum computing. Before founding Azulene, he led chemistry-focused research at Intel Labs. He was drawn to quantum computing because it promised to solve chemistry problems with extreme precision. But when it became clear the hardware might still be decades away, Sawaya looked for faster solutions — and Azulene was the answer.
“At the moment, to really know anything for sure in chemistry, you really have to run the laboratory experiment,” he explains. “But in the macroscopic world, that’s not necessarily true. The example I like to give is the Burj Khalifa, the tallest building in the world. We knew exactly what that building was going to look like before it was even built. Chemistry just isn’t like that.”
That gap between simulation and reality is precisely what Azulene is trying to close. It’s not just about better technology; they’ve rethought everything from the ground up. To support their compute-intensive workloads, the team built their own high-performance cluster in Nevada, cutting costs by nearly 85% compared to using cloud providers like AWS. This practical decision reflects Azulene’s hands-on approach to problem-solving.
That same mindset carried into their experience at Bakar Bio Labs. Since joining the Community Access Bakar Labs (CABL) program, the team has become part of a community of scientists and entrepreneurs tackling complex, technical challenges every day. “There are good vibes there,” Sawaya says. “It’s hard to create that. I’m impressed at how well they pulled that off.”
That collaborative energy, combined with Azulene’s rigorous approach, has fueled the company’s momentum. With a growing library of high-quality data and steadily improving models, the team is getting closer to its goal of making molecular simulation as dependable and predictable as engineering design.
In a field long defined by uncertainty, Azulene is proving that chemistry doesn’t have to be guesswork: it can be computed.
