cloud architecture ml for predictive analytics

2 minutes

While predictive analysis has been around for decades, interactive and easy-to-use software solutions mean that it no longer requires mathematicians and statisticians to perform.

Predictive analytics describes the use of data, statistical algorithms and machine learning techniques to identify future outcomes based on historical data. A valuable capability in every sector, more and more organisations are utilising predictive analysis to increase the bottom line and commercial advantage.

Predictive models allow companies to forecast and manage resources for more efficient operations as super material manufacturer Element 6 (E6) discovered.


The high–end innovator, researcher and manufacturer of advanced super-materials recognised the value of incorporating data science and machine learning as a way to realise efficiencies in its innovation cycle through greater insight and automation. E6 engaged expertise from T-DAB.AI (T-DAB) to design the cloud-hosted architecture for a machine learning and predictive analytics engine. They needed a solution that could handle multiple data inputs and formats, consider a hybrid approach of both streaming and batch processes, manage the edge deployment of predictive algorithms and scale to support a global deployment. The solution would also need to utilise a bespoke, automated ETL process previously developed with the E6 team and analyse the data before being used to train machine learning algorithms to support the manufacturer’s innovation challenges.


Working closely with E6’s team to understand the business and technical, T-DAB designed and developed a cloud architecture necessary to support their data tooling. Through a cloud & data strategy session, T-DAB and E6 outlined the solution should deliver an IoT pipeline to the cloud hosted architecture. The architecture would ingest, transform, store, and analyse data, manage the edge deployment of predictive algorithms (i.e. deep learning prediction of failure) and integrate with a user interface to provide results back to the operators. A proof of concept solution was firstly designed for AWS and then re-designed in Microsoft Azure as a strategic vendor.


Since its launch, the innovative, bespoke solution from T-DAB has already led to novel insights into material properties as well as performance. The system architecture carries out automated batch ingestion from E6’s filing system, while analytics are regularly carried out using the R Server instance and the MS SQL database. The server connects to provide an analytics layer to enable training machine learning regression-like algorithms to predict future super material performance with accuracy. Now, the client is able to carry out testing in a systematic and holistic manner. Most importantly, it’s enabled E6 to visualise and analyse the results of testing as a whole, generating greater efficiencies.