AWS and HighByte: Untangling Industrial Data in the Age of AI

As manufacturers move Industrial AI from experimentation to production, the need for contextualized, ready-to-use industrial data is becoming harder to ignore.

Manufacturers are looking beyond isolated pilots toward applications that can improve maintenance, quality, throughput, and decision-making across operations. But industrial data is rarely ready for those use cases as it arrives from the edge.

HighByte and Amazon Web Services (AWS) address this challenge with a collaboration that bridges the edge and the cloud, effectively closing the OT-IT gap. HighByte Intelligence Hub prepares industrial data close to the source, while AWS provides the services to deploy advanced analytics and AI agents at scale.

The industrial data challenge

Messy industrial data isn’t a new challenge. Industrial environments contain a mix of equipment, controllers, sensors, SCADA systems, historians, databases, files, and business applications, often built over decades of investment. A lot of the data is spread across systems and separated from the information needed to interpret it. Because of that, a raw tag could make sense to the engineer who knows the machine but might mean very little to a data scientist or a consuming application that lives off premises.

AI changes the scale of the data problem. As more applications, models, and agents rely on industrial data, manufacturers need a repeatable way to prepare and deliver it across systems and sites—which starts with an Industrial DataOps foundation. 

Industrial DataOps: Adding critical context

HighByte Intelligence Hub connects industrial and enterprise data sources and adds structure and context to industrial data. It can apply models, set metadata, and create schema so data is prepared for downstream applications and cloud consumption.

This includes telemetry, transactional, historical, and file data, enriched with context from systems focused on maintenance, quality, inventory, scheduling, and more.

Different applications may need different views of that data, but those views can build on the same underlying context rather than starting from the raw data each time. That contextualized data then becomes the foundation for how the Intelligence Hub works with AWS services.

HighByte x AWS

The AWS Industrial Data Fabric brings together AWS services and partner technologies that enable manufacturers to access contextualized industrial data via bulk, streaming, query, or API orchestration from factory floor systems. Depending on the use case, the reference architecture may include AWS IoT SiteWise and Amazon S3 alongside partner technologies like HighByte Intelligence Hub.

Within that architecture, the Intelligence Hub prepares industrial data at the edge before downstream AWS services consume it, store it, and put it to work. Once that foundation is in place, manufacturers can use the data for analytics, machine learning, and AI applications. Furthermore, by applying AI for DataOps, users can leverage Amazon Bedrock to map data tags and build pipelines with natural language in the Intelligence Hub. HighByte Intelligence Hub adds the context required for AI agents to operate reliably in production environments, while generative AI capabilities, including Amazon Bedrock, can be used to build pipelines using natural language.

The architecture is bidirectional. AWS services can serve as both a destination and a source for the Intelligence Hub, allowing it to query analytical data products in AWS and bring those insights back to the edge for use in operational systems.

This gives manufacturers a way to connect operational data with cloud applications without treating every new use case as a separate data integration project. That connection makes the data available—but availability alone doesn't determine what to do with it. This requires a business strategy and a use case. 

Starting with the use case

It can be tempting to move large volumes of raw industrial data to the cloud and decide how to use it later. But a more practical approach is to start with the outcome you want to achieve and work backward from there.
 
For example, predictive maintenance may require combining asset conditions with maintenance records and failure history. Quality applications may require correlating process parameters with product and inspection data. Working backward from the use case helps determine which data matters and what context needs to be added.
 
This becomes increasingly important as the same underlying data supports more downstream applications. An engineer may be able to map tags for one dashboard or combine a few sources for a single analysis. Repeating that work independently across predictive maintenance, analytics, digital twins, and AI applications becomes much harder to manage.

How Gousto built a foundation for AI

Gousto, a UK recipe-box manufacturer, shows what this can look like in practice.

The company started with a specific goal: reducing repair time in a complex packaging environment. With HighByte and AWS, the company built an industrial data foundation that helped transition maintenance from reactive to preventive and eventually to prescriptive with operational simulations.

Within six months, OEE increased from 70% to 95%, and mean time to repair fell to less than 15 minutes.

Since then, Gousto has continued to build on that foundation. In a recent factory automation project, the company used the Intelligence Hub to transform raw operational data from robot arm PLCs into structured, contextualized models before sending it to AWS IoT Core and ultimately to its warehouse systems. The architecture was designed to keep the robots and factory systems synchronized in near real time, targeting less than 500 milliseconds between a box being palletized and the corresponding update reaching the Warehouse Control System.

What began as a focused effort to improve maintenance became a broader data foundation that Gousto could extend to new automation initiatives. That ability to build on the same foundation, rather than starting over for each project, is what makes this approach valuable as manufacturers expand into analytics, automation, and AI.

Building for AI

AI did not create the industrial data contextualization problem. But it does increase the number of applications, models, and agents that need to understand that data. It also increases the pressure to move from a use case to production more quickly, making a consistent approach to context even more important.

HighByte and AWS help manufacturers prepare industrial data at the edge and make it available in the cloud for analytics and AI. That creates a consistent foundation they can build on as new applications emerge.