PRISM's methods were built to predict how patients decline, turning a messy, gap-ridden medical record into a scientific instrument. A pumping station turns out to be not so different. It has a homeostasis, it holds it under load, and it loses it before it dies. That is what we learned to read.
There are a thousand unknowns in why a kidney fails, and clinicians still see it coming. Not from any single lab value crossing a line, but from the pattern across many of them shifting together. That is the insight PRISM is built on, and it is why threshold alarms are the wrong instrument for infrastructure.
The techniques came with it. Embedding a record into a high-dimensional space so you can do arithmetic on behavior rather than eyeball a trend line. Picking a story up mid-stream after the subject has been off the grid for a year. Our specialty is data that arrives late, out of order, or not at all. Industrial telemetry looks like that, and clinical records look worse. Working in that condition, rather than requiring a cleanup project first, is the difference between a demo and a deployment.
Firstlook is headquartered in Dallas and builds entirely in the United States.
“Many of the biological principles I was using for the medical record turned out to apply directly to mechanical systems. Homeostasis, the tendency of a system to hold a steady state, being the first one.”
Brody Holohan, PhDChief Technology OfficerPrediction models deployed at national scale in oncology, and a research career spent turning unusable records into signal.
Built and scaled a services business to 250+ Global 1000 clients before pointing the same discipline at infrastructure.
We are deliberately not naming counterparties here. Each of these is a real relationship at the stage described, and we will be specific on a call under NDA.
Prediction delivered through an OEM's installed base rather than sold site by site. It reaches assets that per-unit hardware economics will never justify monitoring.
5,000 installations, 40 statesThe private-equity-backed operator behind the validated deployment, now expanding from the original 391-asset footprint to nationwide well monitoring and leak detection, still with zero field hardware installed.
391 assets → nationwideMarkets where failure costs include regulatory penalties and service commitments, not only repairs. That is where months of advance notice are worth the most.
Two regions in discussionWe are also in partnership discussions with a data center operations firm founded by former hyperscaler facility-operations leadership. The aim is to put PRISM in front of the operators whose buildings they already run. That one is exploratory, and we would rather say so than imply a signed channel. Full reference list available on request
Clinical machine learning, enterprise operations and diagnostics, pointed at the systems most teams treat as someone else's problem.
Built a previous venture to 250+ Global 1000 clients. PhD, Northwestern.
Led clinical prediction models deployed across the US oncology system. MBA, Kellogg.
Postdoctoral researcher at Yale. PhD, UT Southwestern.
Architect of PRISM and its Feature Intelligence Layer. PhD, UT Southwestern.
No write path to plant controls exists in the product, the evaluation runs entirely offline on flat files under NDA, and we state our SOC 2 and ISO 27001 position in writing rather than on a badge.
Bring your most skeptical engineer to the call. The retrospective is built to withstand that scrutiny.