Kelvin Platform Overview¶

Video Tutorial¶

You can choose to watch the video walkthrough of this tutorial, or continue below to follow the step-by-step written guide.

What is Kelvin?¶

Kelvin is an intelligent automation platform for industrial operations.

Your PLCs, DCS, and RTUs already handle raw control at the asset level. Kelvin sits above them as the intelligence layer, turning the data they generate into smarter, higher-level decisions and automated actions across your operation.

What Kelvin enables:

• ML-driven insights from real-time asset data
• Automated decisions across single or multiple assets
• Human-in-the-middle control where manual approval is needed
• Operations-built automation without programming knowledge

Infrastructure¶

Kelvin is made up of three layers.

Users¶

The people who interact with Kelvin fall into three roles:

• Developers and Data Scientists — build applications and machine learning models
• Operations Teams — manage assets and act on automation through the Kelvin UI
• Platform Administrators — configure and maintain the platform infrastructure

Cloud Services¶

The Kelvin cloud hosts the core platform. This includes:

• Databases for asset data, telemetry, logs, and platform data
• The Kelvin UI for operations teams
• APIs and SDKs for developers and data scientists

Edge Systems¶

The edge is the compute layer between your existing data sources and the Kelvin cloud.

An edge system can be any device capable of running Linux and a containerized environment:

• A virtual machine in your data center or cloud
• A physical machine in a control room
• A device located directly at the asset
• A low-power ARM-based computer

The edge connects to your existing data sources such as SCADA systems, process historians, and IIoT devices. It reads data from those systems and processes it in real time.

Platform Components¶

Assets¶

An asset is anything you define it to be.

A beam pump, a motor, a compressor, a truck — if it is relevant to your operations, it is an asset in Kelvin. Everything in the platform connects back to assets.

Data Streams¶

Each asset has data streams — the individual measurements and signals coming from that asset.

Examples include temperature, pressure, flow rate, and speed. You define which streams matter for each asset.

Connectors¶

Connectors pull data from your existing sources into Kelvin.

Each connector links to a data source such as a SCADA system or process controller and feeds the relevant streams into the platform.

Applications¶

Applications, which are also known as Kelvin SmartApps™, is the intelligence that powers your automation.

It is not designed to replace your PLC's, DCS or RTU's. It runs on a layer above them that can provide either or together;

• Higher level intelligence to your automation, such as AI integration
• Bring better automation coordination between all your existing control systems
• Connect your systems and data to third party resources such as your ERP or MES enterprise systems or analytical resources such as Databricks, Amazon EMR, Google BigQuery, etc.

Applications mainly create actions for your team to review and approve.

In Kelvin, this is called a recommendation. It includes the action to take and the supporting evidence, such as graphs, images, text, or third-party links.

You can set actions to wait for approval, or run automatically.

They can also do much more based on your operational needs.

For example, they can process and export data to third-party apps, send alerts and reports, and receive actions from tools such as Slack.

Basically anything you can program, you can do.

Clusters¶

Applications and connectors run on clusters.

A cluster is a managed group of computing resources running on your edge system. It automatically keeps your applications organised, resilient, and scalable.

Kelvin UI¶

The Kelvin UI is the single interface for operations teams to manage everything — assets, data, applications, and actions.

Recommendations¶

In Kelvin, many workflows start as review-and-approve actions.

This core feature is called the Recommendations system.

When an application or task detects a condition, it creates a recommendation with:

• One or more actions to execute, such as changing setpoints, sending notifications, or sending data to third-party systems
• Evidence that explains why, such as graphs, images, text, or links

Teams can choose whether actions run automatically or wait for manual approval.

Actions¶

The most common action is a control change — updating a setpoint directly on an asset.

But actions can be anything:

• Notifications via email, Slack, or other channels
• Data pushed to ERP or manufacturing systems
• Any custom routine you define

Guardrails and Data Quality¶

Kelvin has two built-in safety layers: data quality and guardrails.

• Data quality checks make sure incoming data is reliable before it is used for decisions.
• Guardrails prevent obviously wrong values from being written back to assets, helping avoid significant asset malfunction caused by clearly incorrect write-back actions.

Who Uses Kelvin¶

Developers and Data Scientists¶

Developers build automation programs called Kelvin SmartApps™.

SmartApps are Python-based applications built with the Kelvin SDK. The SDK handles all connectivity to the platform, so developers can focus on the logic.

SmartApps can range from simple rule-based programs to full machine learning pipelines processing real-time data.

For advanced use cases, Kelvin also supports Docker — allowing developers to work in any programming language and connect through the Kelvin API.

Operations Teams¶

Operations teams use the Kelvin UI to:

• Monitor assets and review its data
• Manage and deploy applications to monitor and control their assets
• Act on recommendations — choosing automatic or manual approval per workflow

Operations teams can also build their own automation without any programming knowledge using Kelvin Tasks. Tasks are AI-powered and allow operations to define, deploy, and run new automated routines directly in the UI in minutes.

Platform Administrators¶

Platform administrators set up and maintain everything behind the scenes.

Platform setup:

• Define asset names and types
• Configure data streams and connectors
• Set up and scale clusters

Ongoing management:

• Monitor infrastructure health
• Manage users and access

Data quality configuration:

Administrators define what good data looks like for the platform. This includes monitors for:

• Missing data
• Duplicate data
• Outliers and out-of-range values
• Availability gaps
• Any custom quality checks relevant to your operations

Combined with guardrails on what can be written back to assets, administrators are the last line of defense for safe and accurate platform operation.

Summary¶

By this stage you should have a good idea of what Kelvin offers and the types of Users that can benefit from Kelvin.

Where do you fit in ?

From here you can move to the relevant areas of the Kelvin Academy to learn more about what interest you directly.

Good luck and Happy Learning !

Test Your Knowledge¶

How Did You Do?¶