Time series foundation models for sensor data

Reduce the time for each new use case for all assets in your fleet from months to weeks (and soon to days), using flexible and reusable AI models.

Foundation Model for time-series data

The Problem

Models for each use case take 4 to 6 months to build.

Current AI solutions are built by looking at patterns represented in well-annotated historical data and using feature engineering to build specific features for specific use cases.

The model that is produced is tailored to a specific use case – and can hardly be reused for another, even a similar use case.

However, there are major problems with the current approach.

Lack of well-annotated data

AI relies on annotated data, which is one of the biggest bottlenecks for AI adoption – collecting data is hard, but collecting and then annotating the data is even harder.

Long model build time​

Currently, a model is produced for a specific use case, and can hardly be reused for another, even a similar use case. This makes comprehensive coverage of assets a very long and expensive process.

Blackbox AI that can't be explained

Without interpretability, the output of AI solutions in the engineering domain is untrustworthy. And there is a mistrust of AI adoption for fleet services.

One model. Multiple use cases.

Amygda’s flexible and reusable AI models reduce the time for each new use case from months to weeks.

A shift towards flexible and reusable AI models. Amygda’s time series foundation models can be used for multiple use cases including smart maintenance, route optimisation, or even equipment life forecasting.

This means companies don’t have to wait for large amounts of data to be annotated in order for them to start seeing returns on their investment in AI.

By not relying on well-annotated historical information or labels, we remove the major blocker in scaling AI across your whole fleet of multiple use cases, enabling full fleet coverage for any equipment.

Amygda Foundation Model infrastructure

Benefits of this new approach

Amygda’s AI models overcome a major challenge in the lack of large sets of well-annotated data. This presents a significant advantage in real-world applications, as it eliminates the risk of equipment failure on critical assets.

Amygda’s models are built using unsupervised and semi-supervised techniques, with further development focusing on self-supervised learning. This approach has reduced reliance on well-annotated datasets, paving the way to addressing novel, diverse and less understood use cases.

AI projects consume a long time in data engineering and rebuilding each time from scratch. By removing this rebuild we can drastically speed up time to production.

Amygda’s models are built with a focus on transfer learning. This has enabled reusability as Amygda’s time series foundation models can be used for multiple use cases, including smart maintenance, route optimisation, and equipment life forecasting. This approach has sped up data centric aspects from months to days, with a vision to reduce the time further to minutes.

The future of work in industries involves the seamless integration of human and AI efforts. The success of this collaboration relies upon the ability of AI to provide transparent and reliable decision-making processes.

Just as we prefer working with honest and trustworthy colleagues, collaboration with AI requires a similar level of trust. Explainable and Trustworthy AI is a prerequisite for any successful embedding of AI in operational processes.

At Amygda, solutions are built to not only provide an interpretable output but also deliver insights into the factors and reasoning behind the decision making process. This is done by the use of explainability algorithms that sit on top of our underlying solutions.

Why now?

By 2025, enterprise AI will be entirely run on AI models that can solve multiple use cases. These models will be general-purpose, meaning that they can be trained on data to learn how to represent it and then fine-tuned for specific tasks. The infrastructure necessary to operationalize these models for multiple data volumes, velocities and veracities is a major innovation.

Why work with us?

Get ROI across your whole fleet

As a fleet operator, you own multiple assets and equipment from several different OEMs. Our AI platform works across any equipment in your fleet, irrespective of its type or age. So you can get whole fleet coverage for any use case.

Embed AI across the business

Scaling AI improves productivity. Multiple teams managing fleets of assets can now benefit from Amygda's solutions. Our AI platform is scalable and enables business leaders to embed AI in various use cases across the business.

Faster time to value

Equipped with billions of data points from 100s of different equipment types, Amygda is already demonstrating a reduction in time to value from months to weeks, and in some instances, days. So you can demonstrate business value faster.

"We are at pivotal moment when cross industry developments in AI can finally be effectively leveraged in the very traditional transport domains where the application has been slow, painful and impractical. It requires a whole new way of thinking and Amygda's reusable, flexible and explainable AI changes how we deliver the promise of AI to the transport industry. The shift has already begun and Amygda's leading the way!"
Shaheryar Khan
Co-Founder and AI Lead

Download our White Paper