Filip
Auksztol

Quantum Boost is a small company with an enormous grudge: against guesswork. Auksztol founded it in 2021 to do something chemists have wanted forever and rarely had - design the next experiment instead of stumbling into it. The platform is, in the company's own words, "your own AI formulation assistant," a system that watches the results of each lab experiment and tells you, with a statistician's nerve, what to mix next.

The pitch is unglamorous and exact. Take a chemist developing an ink, a coating, a drug formulation, a specialty polymer. They have dozens of dials - concentrations, additives, temperatures, ratios - and a near-infinite space of combinations. The old way is intuition plus a lot of failed beakers. Quantum Boost's way is Bayesian optimization: a model that learns from every data point and steers the next experiment toward the answer. The claim that follows is bold and specific - formulation development time cut by 50 to 80 percent, and fewer wasted experiments along the way.

It is the kind of promise that only sounds reasonable coming from someone who has spent years thinking about systems most people will never see. Auksztol is, by training, a quantum physicist. Before he was selling chemists a better way to run their benches, he was building qubits out of cold atoms.

That is the strange, specific fact at the center of his story. The math that now recommends your next paint experiment grew out of a doctorate spent shaping light into traps for matter that behaves like a single quantum wave.

There is a temptation to treat the jump from quantum optics to chemistry R&D as a swerve. It isn't. The discipline that defines both is experimental design - the art of learning the most from the fewest trials. A physicist who has spent years coaxing fragile quantum states into existence understands, viscerally, how expensive a single experiment can be and how much information a good model can wring from each one.

That is the bet underneath Quantum Boost. Chemists, brilliant as they are, often run experiments the way they were taught: one variable at a time, intuition first, a binder of past results that no algorithm ever reads. Auksztol's platform proposes a quiet inversion. Let the data decide where to look. Let a Bayesian model hold the whole response surface in its head and point at the next most promising mixture. The chemist stays the expert; the software handles the combinatorics that no human brain was built to navigate.

The economics are the seductive part. In coatings, in pharma, in specialty chemicals, time-to-formulation is the cost that dwarfs the others. Shave 50 to 80 percent off that, and you have not just saved money - you have changed what a small team can attempt. A research group that could afford to explore three formulations can suddenly afford thirty. That is the real product: not speed for its own sake, but ambition that used to be unaffordable.

It helps that the founder is fluent in the language of skeptics. Auksztol's academic record - peer-reviewed papers, a doctorate, a couple hundred citations - is not decoration. It is credibility with exactly the audience that matters: the technical R&D leads who have seen a hundred "AI will fix your lab" pitches and reflexively distrust them. When the person making the claim has himself sweated over reproducibility, error bars, and the cruelty of real data, the pitch lands differently.

None of this is a finished story. Quantum Boost is small, early, and operating in a market - applying machine learning to materials discovery - that is suddenly crowded with ambition. The company's reported funding has been modest, its headcount lean. But the shape of the wager is clear and unusually honest. It does not promise to replace chemists. It promises to stop wasting their time.

And that, in the end, is the most quietly radical thing about Filip Auksztol. He left a field built on elegant, abstract precision to attack a problem most people find tediously concrete. He decided that the romance was not in the quantum, but in the question: how do you discover something new without burning a year to find it? His answer is a piece of software and a stubborn belief that guesswork had it coming.