Ai-OPs

Ai-OPs is an industrial AI company based in Mobile, Alabama, that develops software for deploying machine learning directly into plant control environments. Its offering centers on two products: Ronin, an application in which engineers and data scientists build, train and maintain ML models tailored to industrial processes, described as reaching full production by 2026; and Koios, a self-hosted inferencing platform that runs trained models alongside DCS, PLC and SCADA systems using standard industrial protocols, with support for air-gapped, real-time and hardware-agnostic deployments.

The platform covers data collection, real-time AI inference on live process data, and operational visualization. Documented capabilities include connectivity to OPC-UA, Modbus TCP, Ethernet/IP, SOAP, SQL and REST endpoints; deployment on-prem, in cloud or air-gapped environments via Docker or Kubernetes; custom logic such as PID controllers, state machines and data transforms written in Python and deployed as hot-reloadable packages; real-time dashboards, historical trending and process KPIs; and enterprise controls including role-based access, encrypted connections and audit trails. The company positions the combination of Ronin and Koios as enabling closed-loop control using deep reinforcement learning and other ML techniques, interfacing directly with production hardware to stabilize systems, reduce valve cycling and improve yields.

Koios 1.2, released as generally available on September 16, 2026, added a three-pane workspace for building components, isolated package stacks tied to user-uploaded wheels, model inputs that can be carved into named sub-inputs with individual window lengths and sample rates, access to serial, GPIO, I2C and SPI host devices without privileged containers, an append-only audit log with its own retention and export plus TLS certificate installation, and streaming trend exports.

Founding story

Ai-OPs was founded in Mobile, Alabama by engineers Phillip Hansel, Curtis Wright, Ryan Hutchison and Connor Smith. The founders had spent years working in industrial control systems and observed that plants generated large volumes of data without a straightforward, scalable way to apply it, with engineering teams stretched thin and upgrades expensive. The company formed around closing that deployment gap between available optimization techniques and production hardware.

Business model

Ai-OPs licenses software deployed in customer environments rather than operating a hosted service: Koios is distributed as a container image (a Docker pull of aiopinc/koios is promoted on the company site) and runs on-prem, in cloud or air-gapped networks under the customer's control, with Ronin used by customer engineering teams to build and own their own models.

Traction

The company reports more than 30,000 hours of production runtime, an average 600% return on investment and under 1% added management overhead. At an early industrial installation, Koios was credited with stabilizing temperature fluctuations, reducing valve cycling and extending run times between maintenance intervals; a Sunoco LP transmix manager describes similar results on a hydrotreater process unit. As of May 2025 the team was expanding with new installations in motion.

Latest developments

Koios 1.2.0 reached general availability on September 16, 2026, adding a unified component-building workspace, isolated package stacks, differently-sampled model inputs, non-privileged serial/GPIO/I2C/SPI device access, an append-only audit log with certificate management, and streaming trend exports; OPC-UA support moves onto the first full release of its client library.