Utilities such as water and electricity are responsible for some of the most critical and costly infrastructure in society. They collect enormous amounts of information from pumps, valves, sensors and other assets, but much of that data remains isolated within their facilities—often separated by cybersecurity, privacy and regulatory constraints. In many cases, there is no centralized intelligence capable of bringing those inputs together and understanding what they mean in the context of an entire operation.
That is the problem John Armin, founder of Utilyst (short for “Utility analyst”) aims to solve. The startup is developing what it calls a “Utility Brain”—a system that combines real-time sensor data with historical records, maintenance logs, operating procedures and the knowledge of experienced operators.
On top of all this sits a small language model capable of interpreting the information. The goal is ultimately to give utilities a layer of intelligence that can understand how their assets are behaving and help identify problems before they become costly failures.
“We can bring intelligence that is missed in many of these utilities,” Armin said. “Our goal is to capture operational knowledge in these models.”
Capture the data, then analyze
Utilyst begins by working with a utility to identify and connect the data that utility already collects, mapping the “tags” and sensors attached to an asset and establishing a pipeline to bring that information into the Utilyst system.
Once connected, Utilyst deploys its hardware on the utility’s premises and connects it to the asset’s live data. The startup also incorporates maintenance records, operational logs and standard operating procedures. Experienced operators are then interviewed to evaluate the model’s recommendations.
“The operators see the models running, see how the model digests sensor data and proposes suggestions,” Armin said. “Then those experienced operators approve or disapprove [the model’s suggestions], and that feedback loop is completed between what we train and predict and what the operators evaluate and approve.”
Utilyst’s analysis architecture has three basic pieces: data integration, predictive models and a small language model. The data integration layer brings in information from the utility’s different sources, while machine-learning models analyze that data to make predictions about an asset, such as the possibility of a failure.
Utilyst also uses causal modeling to examine how changes in one asset can affect upstream and downstream parts of a system, which Armin said is a major differentiator.
“The language model is the engine,” Armin said. “It digests real-time data, predictions from the trained machine-learning models, and reviews the historical records—and, based on that, suggests procedures on how to operate the specific asset. But a human is always the final decision-maker. The Utility Brain’s recommendations don’t interact with the asset directly. ”
Everyday operations
Utilyst presents operators with information through a dashboard built around the asset. The platform can provide a 3D view of the equipment and its connected sensors while displaying real-time data, model predictions and recommendations from the language model. Operators, in turn, can interact directly with the system to ask questions about what they are seeing or why a particular recommendation was made.
If the system identifies a potential problem or recommends an action, the operator can review the information and decide whether to follow, modify or reject the recommendation. That decision then becomes part of the feedback loop used to refine the system.
“Generally speaking, the main value comes from understanding the asset better, in terms of what is happening, and having the capabilities to look at so many variables at the same time, which usually for humans is not possible,” Armin said.
Compliance
Utilities often make for slow-moving and highly bureaucratic organizations, meaning a new system needs to offer a clear reason to justify the time and effort required of adoption. For Utilyst, that reason is compliance. The platform is designed to help operators continuously monitor whether an asset is operating within required parameters and flag conditions that could put the utility out of compliance.
Armin said efficiency and cost savings are additional benefits—through reduced energy or chemical use, improved maintenance or less waste—but compliance comes first.
“The main thing that they want to see from an operator’s perspective is that, ‘are we in compliance with what we do?’” Armin said.
A serious failure at a water utility, for example, could lead to a boil-water advisory and damage public confidence. By continuously analyzing the conditions surrounding an asset and bringing potential problems to an operator’s attention, Utilyst aims to help utilities identify issues before they become compliance problems.
Traction and the road ahead
Utilyst is still in the early stages of deploying its technology, but the startup has begun moving from development into commercial use. It currently has one paying customer, a large East Coast utility, while another is finalizing a contract.
Beyond its customers, Utilyst has been selected for several competitive startup and industry programs. It joined Techstars Alabama EnergyTech as one of eight companies chosen from more than 400 applicants and was one of five selected from 102 applicants for Black & Veatch’s IgniteX program. Closer to home in North Carolina, Utilyst was a finalist for a SEED grant from NC IDEA, completed the Launch Powered by KPMG accelerator as part of its third cohort, and pitched on stage at Grep-a-Palooza 2026.
QUICK BITS
Startup: Utilyst
Founder: John Armin
Founded: 2025
Team size: 4
Location: Raleigh
Website: utilyst.com
Funding: Bootstrapped
For now, Armin said the company is focused on expanding within its two primary markets: water and electricity. Over the next year, the goal is to add more utilities, prove the value of the system across different applications and develop repeatable deployments that can be adapted from one utility to another.
Longer term, Armin envisions Utilyst becoming more of a self-service platform that utilities could deploy on their own hardware with minimal setup. He also sees the technology eventually expanding into industries such as oil and gas and other industrial facilities.
The broader goal is to turn what has begun as a highly customized system for individual assets into a platform capable of serving a much wider range of infrastructure.
