Who we are, and why we think the surfaces of a running line should not be dark.
Lumina builds the continuous record of a food plant. An operations platform, live today, that turns every sample, form, reading and finding into a record with an owner, a clock and a citation. And a proprietary line-side sensor, lab-demonstrated and in build, to read a surface where contamination actually happens, instead of waiting on a lab that answers days later.
The founding thesis is simple. Today's toolkit measures the wrong thing, or measures it too late. A confirmed culture result takes four to six days. PCR is faster but still off the floor and still batch-scheduled. ATP is instant, but it reads organic residue, so it cannot tell you whether what it found is a pathogen. Bench-top Raman has the specificity and none of the reach. And a swab samples one point at one moment, so in between, the surfaces of a running line are effectively dark. That interval is what continuous measurement is being built to close.
The stakes are not abstract. There are roughly 1,600 listeriosis cases and 260 deaths a year in the United States, almost all traced to contaminated ready-to-eat products. That is where we started. The ambition is a food system where that number is a historical statistic rather than an annual one.
The line keeps running while the culture plate incubates.
The team
We are engineers with deep expertise in remote sensing, machine vision, and data analysis. Our founder spent thirteen years building sensor networks at scale, most recently a public-safety sensor network that passed 150,000 devices: always on, unattended, and judged on uptime as much as accuracy. The doctrine carries over: read the product and the process, never the person.
Factories need the same capabilities. Production lines run around the clock, conditions are difficult to see in real time, and incidents are often reconstructed afterward from paper logs and disconnected systems. Lumina identifies risk as it emerges and automatically creates a reliable record along the way.
13 years
Building sensor networks at scale, judged on uptime.
Seven
Under an exclusive university license, three patents issued.
150,000
Devices on the public-safety network our founder architected.
Atlanta
Georgia, United States. Venture-backed.
How we work
When our system cannot ground an answer in your records, it says so rather than inventing one. In a field where a wrong answer becomes a liability decision, calibrated confidence is not modesty. It is the feature that makes an automated safety system usable at all.
Lumina runs in the cloud, with each customer's data isolated in its own tenant. Your specs, SOPs, HACCP plans and results are used to answer your questions and build your record. They are not used to train third-party models, and the assistant is not a shared wrapper over a general chatbot: it answers only from what you uploaded, and cites the passage. We intend to train Lumina's own models on de-identified data, under the terms of your agreement, and never anyone else's. The plant sets what is captured and how long it is kept.
Lumina keeps the record around the environmental monitoring program you already run: more of the plant covered, more often, by the same team. Confirmatory lab testing stays authoritative, so we are not asking you to take a regulatory risk to try something new.
Talk to us
Today that surface gets swabbed on a schedule, and you find out what was living on it four to six days after the line ran. On the platform, the positive opens its own issue the moment the lab report lands, the corrective action lives in the issue with Ask answering from your own SOPs and the citation attached, and the record of all of it assembles as the work happens rather than the week before an audit. Read that surface continuously and the signature shows up while it is still forming: that is what the sensor is being built to do.
We are ready to move your program from reacting after the fact to preventing while the line runs. Tell us which surface to start with.
