Updated 21 Nov 2022Reading time: 3 mins

Interview with digiLab

digiLab is a data science company based in Exeter. Our founders have been key members of the Data-Centric Engineering team at Exeter University

What is digiLab?

digiLab is a data science technology company based in Exeter. Our founders have been key members of the Data Centric Engineering team at Exeter and have built a comprehensive background in industrial applications of data science. We are now using that expertise to provide high value data insights to our clients through a combination of software products and consulting services.

What type of clients do you work for?

The digiLab solutions can provide significant value for companies in a wide range of industries. A t present we are working closely with leading businesses in the water industry, in nuclear fusion and nuclear decommissioning, and we are looking at some exciting projects in health care.

What is the basis of the digiLab solution?

Fundamentally what we do is probabilistic forecasting. Our solutions examine the client’s data – often time-series or sequential data – and then use the latest in Bayesian machine learning techniques to enhance the understanding of that data and extract valuable insights. But we offer much more than one-off analysis: our software tools use advanced statistics to identify the areas of data with the greatest uncertainty, and then make recommendations for prioritising new data sampling to improve the data even further.  This capability enables businesses to use enhanced data to make more informed decisions. We call it decision intelligence.

How does that differ from other analysis solutions?

We’ve developed a unique combination of capabilities in Data Cleaning, Uncertainty Quantification and Decision Intelligence, based on advanced statistical methods and machine learning. We’re making that technology available to our clients through an open platform that is highly interoperable, connecting to other simulation and analysis tools. This is much more than traditional data analysis methods that are limited to one-off assessments and that struggle with inconsistent and incomplete data.

What types of problem can be addressed with the digiLab solution?

We are uniquely placed to address problems where the data is incomplete, with gaps in time series or missing locations, and where the data includes duplicate and redundant information. An example is the data coming from sensors that measure water quality in a treatment plant. Some sensors may be offline, some sending data more frequently than others, so the aggregated data is messy and very difficult to interpret. Our solutions can be used to clean the data, and then use machine learning methods to predict the dosage of additives that will maximise the water quality, while minimising the amount and cost of the additives. A combination of enhanced understanding of the data followed by more accurate predictions of the outcomes from specific decisions. Another example is in sampling of radioactivity in preparation for decommissioning of nuclear power facilities. Each test is expensive and the current approaches lead to more samples than are necessary, but not necessarily in the best locations. The digiLab solutions can be used to predict the most effective sampling locations, minimising the cost of testing and maximising the accuracy of the derived data.

Why is this becoming more important now?

We are seeing a very substantial increase in the number of connected devices and sensors across many areas of society and industry, and a corresponding increase in the amount of data those devices are producing. Unfortunately, much of that data is messy, inconsistent and incomplete, and in its raw form is simply not suitable for supporting critical business decisions. With the development of machine learning combined with advanced statistics, the digLab solutions allow companies to clean their data, analyse and characterise it, understand the levels of confidence and uncertainty, and then apply predictive methods to support their decision-making. The digiLab solutions allow business to unlock the true value of their data.

This sounds very useful, but extremely complicated, so how do your clients get the value from the solution when they may not be experts in computational data science?

The digiLab team has created user-friendly interfaces that allow non-experts to use the advanced statistical methods and artificial intelligence approaches that the digiLab experts have developed. The approach of using an open platform also allows the digiLab solutions to be combined with a wide range of existing analysis and simulation techniques, including FEA, CFD, and associated design-of-experiments and optmisiation tools. DigiLab also provides consulting services, training and support, to help clients deploy the solutions and gain maximum value from the technology.

In summary, how will digiLab contribute to your clients business success?

We believe there are very many companies that have large, complicated datasets, often unstructured and messy. They know there is valuable insight to be gained from the data, but they do not know how to exploit what they have.  At digiLab, we can work with those clients to help them clean their data and deploy the very latest AI algorithms to provide unprecedented levels of understanding and insight. We can make accurate forecasts and provide what we call decision intelligence. We enable our clients to make better use of their data, to make improved strategic decisions that will help them to be more sustainable, more robust and more profitable.