Profile
The Man Who Taught Plasma to Listen
There is a moment in every battery factory that nobody talks about. The electrode comes off the coating line. It looks fine. The line supervisor says it looks fine. And then, somewhere downstream - in the pack, in the car, in the field - it turns out it was not fine. Jared O'Leary built a machine for that moment.
PlasmaSens, the platform his company SirenOpt has spent three years engineering, sends a cold atmospheric plasma over a material's surface and reads back 213,000 data points in milliseconds. Thickness. Density. Resistivity. Chemical composition. All at once. All without cutting, probing, or destroying the sample. The measurement happens at the speed of a production line, which is the only speed that matters in manufacturing.
O'Leary is not the kind of founder who discovered manufacturing from the outside. He grew up in it intellectually, earning a B.S. with honors and distinction in Chemical Engineering from Stanford, then a Ph.D. from UC Berkeley under Professor Ali Mesbah - the same researcher who co-founded SirenOpt with him when the spinout came together in August 2022. His doctoral work centered on learning-based methods to characterize, model, and control advanced materials manufacturing processes. The company is, in a direct sense, his dissertation made real.
"Measuring certain types of properties inline or non-destructively that otherwise cannot be measured either non-destructively or in line."
- Jared O'Leary, Chipstrat Chat interviewBetween Stanford and Berkeley, O'Leary spent two and a half years at Theranos as a Systems Integration and Validation Engineer, eventually becoming a Team Lead. He left in mid-2016, returned to academia, and spent six years doing the kind of rigorous scientific work that produces award-winning papers and defensible technology. That experience - watching what happens when measurement is faked rather than engineered - is not something he has to explain. The contrast is built into the product.
SirenOpt's investors read like a who's-who of industries that desperately need better process control: Hitachi Ventures (industrial automation, sensing), InMotion Ventures (JLR's venture arm, which needs battery quality to hit their EV targets), Voyager Ventures, and Visionaries Tomorrow. The $6.5 million strategic round that closed in late 2025 followed a $6.6 million seed in July 2024, bringing the total to $16.1 million. Separately, the California Energy Commission chipped in $2.4 million in BRIDGE funding specifically for battery electrode manufacturing - a signal that state-level green energy policy is now writing checks to the people solving the factory floor problem, not just the chemistry lab problem.
Physics First. Neural Networks Second.
O'Leary is vocal about one thing that separates SirenOpt's AI from the current wave of machine learning enthusiasm in manufacturing: physics-informed models over black-box neural networks. In the Chipstrat Chat interview and elsewhere, he has been consistent that for high-stakes, high-precision manufacturing decisions, you need a model that respects physical laws - not one that is trying to pattern-match its way to an answer with no understanding of what the numbers mean.
This is not a minor philosophical preference. It is the architecture of the product. PlasmaSens pairs plasma spectroscopy with physics-informed deep learning in a way that means the measurements are interpretable, traceable, and correct in the ways that matter to a process engineer trying to understand why a batch of battery electrodes failed a specification.
The O. Hugo Schuck Best Paper Award at the American Control Conference 2024 - awarded for a paper on physics-informed deep learning approaches to stochastic control of colloidal self-assembly - arrived at roughly the same time SirenOpt was shipping its first alpha units. The award recognized the scientific foundation; the alpha units were the first proof that the foundation could survive contact with a factory.