INSIGHTS:
Portfolio
/

Welcome our latest investment: SirenOpt

Alisha Jani
•
October 5, 2026

The sensing & intelligence layer for advanced manufacturing

Advanced manufacturing forms the backbone of the global economy, contributing ~$8 trillion in direct value every year. That value is made or lost beneath the surface: in the chemistry of a battery electrode, the integrity of a chip bond, or the uniformity of a turbine blade coating. Yet manufacturers have no way to measure subsurface properties, detect drift, or predict performance during production. 

Instead, they approximate. Manufacturers pull parts at random, cut them open, and wait days for a lab to report back on quality. In the meantime, the line runs on, producing tens of millions of dollars of ‘exposed’ product. If the sample fails testing, the exposed batch is scrapped. If the sample passes, all manufacturers can do is cross their fingers and hope it was representative of the batch. 

Offline testing is the weak link in an otherwise advanced industry, and the cost of its imprecision is enormous.. A leading turbine maker has $4B+ in demand that it can’t fill because of frequent testing and rework. Low chip yields reportedly pushed Qualcomm and Google from Samsung to TSMC, costing Samsung over $1B a year. And a major European automaker faced a $4B full vehicle recall over a battery contaminant that offline testing missed. 

The ability to see inside a manufacturing line, to catch and diagnose problems as they happen, would fundamentally shift the efficiency and reliability of our manufacturing base.

SirenOpt is building the missing sensing and intelligence layer for manufacturing. 

Founded in 2022 by Jared O’Leary and Ali Mesbah, SirenOpt has developed the first tool that can resolve multiple material properties, in real time, on the line, allowing manufacturers to pinpoint exactly where and how defects form during production. This process-property visibility lays the foundation for predictive capabilities and, ultimately, for automated factory control. 

"SirenOpt is delivering what many have long considered the holy grail of software-defined metrology." - Jan Marchewski, Hitachi Ventures

‍

Technology

Bringing a new measurement physics to market

Today’s inline sensors are largely optical, so they can only see what light can reach: the surface. The tools that see deeper, like X-ray or cross-sectioning, are slow, low-resolution, or destructive, and most measure only one property at a time.  

SirenOpt closes this gap with a jet of cold atmospheric plasma aimed at the product as it moves down the manufacturing line. In under a millisecond, the plasma’s interaction with the material releases a series of electrical, thermal, and optical emissions. Physics-informed machine learning then translates these signals into properties such as density, resistivity, and chemical composition. 

The hard part, though, is control. Plasma is intensely energetic and can alter what it touches. The jet must be strong enough to produce a clean signal but gentle enough to leave the product unchanged, a balance that is difficult to maintain even on a stationary sample. SirenOpt holds it on roll-to-roll and piece-to-piece production lines - where surfaces are moving, vibrating, and changing shape - by correcting the plasma thousands of times per second. This control system grew out of more than a decade of research at the CTO’s lab at UC Berkeley, and it is the core of SirenOpt’s technology. 

Adopting SirenOpt is deliberately simple. The sensor head mounts above an existing production line, and after a short calibration cycle, the system learns which properties matter most to a given customer. Every reading adds to that customer’s record and over time, the data compounds. SirenOpt can then do more than identify faults: it can predict how a product will perform (inference), trace drift to its source (intelligence) and, ultimately, adjust the line automatically to keep the product within customer specifications (automated process control). Together, these layers form a digital manufacturing stack, built entirely on SirenOpt’s data. 

‍

Unit Economics

The cost of poor quality: $1 at the machine, $100 in the field

The manufacturing industry is used to pricing failure. It’s well understood that a defect costs ~$1 to fix at the machine, $10 loose inside the plant, and $100 once it reaches the field. As the only tool that can measure subsurface characteristics at production speed, SirenOpt is able to isolate defects at the machine, saving manufacturers up to two orders of magnitude in downstream cost. 

This trade-off is what justifies the cost of an instrument, and is why SirenOpt has been met with significant customer demand and strategic investment to-date. 

The hardware sale is just the beginning. Each sensor opens the door to a recurring software relationship that deepens over time: moving from measurement to inference and intelligence (predicting how a product will perform and detecting drifts, respectively), and finally to automated process control. Each layer is monetized separately,  so revenue per account rises with value created. 

We have seen companies create outsized value in metrology before. In the 1990s when semiconductor fabricators shifted manufacturing methods, KLA was the only company with the sensing physics to see critical defects. They took the market, growing revenue twelvefold to $2.1 billion in under a decade. Manufacturing capability had moved past what the dominant sensing modality could perceive, and the company with the better physics dominated the category.

SirenOpt is positioned at the same point, one layer deeper, with a physics moat and data asset that compounds with every sensor head deployed. 

‍

Impact

Building the hardware for the energy transition

SirenOpt's initial markets sit at the core of the energy transition: batteries store energy, turbines generate it, and semiconductors control it. The efficiency savings from scrap, overbuild, and margin across these three industries could avoid as much as 174Mt annual CO2e.  

The larger impact, however, lies in the innovation velocity that visibility unlocks.  Advanced manufacturing production today is causally blind: a change is made, the result arrives later, and nothing initially indicates which variable caused it. This is the main reason battery and semiconductor yields take years to climb. Real-time metrology compresses this feedback loop: every new chemistry, coating and material that follows gets to solution faster, from electrolyzers to solar films and power semiconductors. Against climate change, pace is a critical variable. 

‍

Market Traction

Our conviction in SirenOpt is driven by the shape of their traction. Of the hundreds of prospective customers in conversation and the dozens who have purchased tools or completed initial evaluation, the team has not lost a single account.

Multiple tier-1 battery, turbine, and semiconductor manufacturers have purchased tools and have begun to deploy inline systems. The deepest of these relationships are already exhibiting repeat behavior. Customers are pulling deployment dates forward and placing repeat orders as the system exposes defects that would have otherwise gone unnoticed. To put this into context, expansion across just the customers who deployed systems to-date would represent a multi-billion dollar recurring revenue opportunity.

‍

Founding Team

Jared O'Leary, PhD (CEO), holds a doctorate in chemical engineering from UC Berkeley with a specialization in stochastic and learning-based model predictive control. He spent six years in Mesbah's lab specializing in machine learning for non-equilibrium plasma modelling. 

Prof. Ali Mesbah, PhD (CTO) left his position at UC Berkeley to join SirenOpt full-time. He is one of the foremost experts in model predictive control, non-equilibrium plasma and plasma-surface interaction modelling. 

Every member of Ali’s lab at UC Berkeley followed Jared to SirenOpt. The team is twenty-nine people strong, drawn from across industry (KLA, Onto, Applied Materials) and academia (MIT, Berkeley, UChicago).

‍

Notes from Lead Investor

“Through a new level of mastery on control of cold plasmas, SirenOpt is bringing incredible new abilities to see deeply into materials, helping to drive substantial savings and make the best use of the materials we use across many industries.” - Tom Chi, Managing Partner, At One Ventures
SHARE
No items found.
ALL INSIGHTS
Arrow