The reference point for what comes next.
AI research for real-world signal systems.
Forty years reading formations, steering through the unknown, living in the gap between raw signal and human decision. That's where DATUM comes from.
We stay because the problems are still hard. But the people solving them don't have to come from here.
The same pattern recognition problem shows up in a wearable sensor and an XRF analyzer. The data changes, the structure doesn't. A risk model in genomics and one in geosteering often rely on the same underlying approaches, despite being treated as separate problems. We hire for that kind of thinking, people who recognize the same problem dressed in different data.
That's the edge. That's DATUM.
Independent. Thesis-driven. Building new ways to understand real-world systems.
We work from model to machine.
Models & Research
We develop new approaches to representation and structure in real-world systems. Original architectures, trained for domains where off-the-shelf models break down. Patents in federated learning.
Our work explores multiple directions, including latent state modeling, connectomics, world models, and other emerging approaches. We treat these not as isolated ideas, but as parts of a larger shift in how intelligence is built.
Compute & Infrastructure
The right model means nothing on the wrong system. We design and deploy the infrastructure to run it: on-prem GPU clusters, edge hardware beside physical assets, and hybrid architectures built for environments where cloud assumptions break.
Through our own systems and partnerships with leading compute providers, we build the full stack, not just the model.
Edge Systems
Real systems don't run in ideal conditions. Latency matters. Connectivity fails. Data is incomplete, noisy, delayed.
We build intelligence that holds up under those conditions, running close to the source, making decisions in real time, and degrading gracefully when the system around it doesn't.
Brilliant people solving similar problems in dissimilar industries.
A geologist reading formation signatures and a radiologist reading tissue density are doing the same thing. A federated learning system in energy and one in healthcare rely on the same core principles. We find the people who see those connections, and we put them to work on the hardest versions of those problems.
Not customers. Collaborators.
Researchers, engineers, and organizations working on problems nobody has solved yet. Teams building internal AI capability who need a partner that accelerates rather than competes. Anyone who knows the real work lives between the paper and the production system.
Our first product. Not our last.
Real-time intelligence for drilling operations, built when we applied our research to the domain we know best.
Datum: a fixed reference point from which all other measurements are made. That's what we're building. The point everything else is measured from.