Kairos is Cerasus's physics-grounded, agentic AI/ML platform for the process industry. It fuses historian data with engineering knowledge to predict asset and process health, explain every alert, and let AI agents act within guardrails — with fleet learning across sites that never moves raw plant data.
Generic machine learning on plant data produces alerts nobody trusts. Kairos grounds its models in process physics and equipment knowledge, so a bearing-degradation or dry-gas-seal warning comes with the sensor contributions, the standard it references and the remaining useful life — an explanation an engineer can check.
Around the models sits a command centre: live telemetry, active alerts, an agent and guardrail feed, and an autonomy posture that says exactly what the AI may do on its own. Federated fleet learning improves the models across plants using secure aggregation and differential privacy, so no site's raw data leaves the site.
Plant availability, open alerts, models in production, agent actions and fleet accuracy on one screen with live process telemetry.
Alert queue with severity, remaining useful life, sensor-contribution (SHAP-style) attribution and physics-attributed explanations.
Historian and file onboarding, context/knowledge graph, data-quality governance and demo datasets to start fast.
Physics-plus-ML hybrid models, twin builder and an analyst workbench for exploring signals.
An LLM engine with native function calling runs automation sweeps and recommendations inside a defined autonomy level (L0–L5).
Secure aggregation (FedAvg + proximal), differential privacy, quality-aware update capping and rollback policy across sites.
Connect historians or files; build the context graph; govern data quality.
Train hybrid physics-ML models; validate against history.
Command centre, alerts with explanations, agent actions within guardrails.
Federated updates across the fleet; retrain as plants change.
Screens from the current build. Click any image to open it full size.
ISO 13374Condition monitoring and diagnostics — data processing, communication and presentationISO 17359Condition monitoring and diagnostics of machines — general guidelinesISO 10816 / 20816Mechanical vibration evaluation of machinesISO 22400KPIs for manufacturing operations managementOPC UA (IEC 62541)Data connectivityIEC 62443Security for industrial automation and control systemsISA-18.2Distinguishing predictive warnings from alarmsSend a short brief — plant, service of interest and what you need to achieve — and a Cerasus engineer will respond with a proposed scope and next steps.
Prefer email or phone? sales@cerasus.ai · +974 5506 2743