In the fast-moving multidisciplinary field of commercial data science and engineering, staying abreast and ahead of rapid technological change is key. T-DAB’s mission is to deliver cutting edge solutions incorporating the latest developments in machine and deep learning, and data engineering.
Research and Application
T-DAB bridges the gap between research and application. Our philosophy is to where we can, deliver with proven existing tools. However, our customers typically turn to us to do the things that others cannot. To this extent, we help companies to innovate. This often involves turning to the academic sector to implement the latest developments in the literature and adapting these into commercially viable solutions. Key to our proposition is knowing when to accelerate the customer with existing technology, and when to deliver a cutting-edge solution with bespoke, advanced developments.
Of course, taking the latest developments from research into application comes with an element of risk. Not all of our customers are immediately prepared to take this as part of a project. But they are still looking for functionality that can only be provided by the latest technologies. Others are looking to carry out R&D, but with a partner with a proven track record in this activity. Our customers and partners also want to know that our technical teams are world class, able to not only build, but have the deep understanding that can only come with research.
Recognising this, we set up our T-DAB Lab: Innovation Sandbox in 2019. This is in partnership with Imperial College London who’s engineering, and computer science departments are world class. We offer the opportunity to MSc and MEng students to opportunity to come and work within T-DAB on a research project for their thesis studies. This provides them with a unique opportunity to work within a data science and engineering company, carrying out in-depth research, on real world problems with real data.
Areas of Research
But how do we select our areas of research? The answer here lies in the work that we carry out with our customers. We regularly review the challenges that our customers are facing, the existing technologies for solving these, the gaps that exist, and the trends in the wider research community to understand which fields of research will be most valuable. In 2019, having worked heavily in manufacturing, we elected to work on the latest applications of deep learning to intelligent automation and control problems. In particular, we were interested in advancing our knowledge of how to build deep neural networks acting as digital twins of complex systems, and then how to adapt these using deep reinforcement learning to learn optimal control strategies.
Three students, supervised by a senior data scientist, worked on this problem over a course of 9 months in the context of autopilot control for round the world racing yachts. Taking what we learned, we have successfully delivered this technology into HVAC energy control valve systems for smart buildings and are working on delivering this into can manufacturing production lines.
So, what now in 2020? This year we have expanded the T-DAB Lab in a number of ways. Firstly, we have tripled our domains of research to cover deep learning for control systems, deep learning for natural language processing and text summarisation, deep learning for time to event prediction and multistep forecasting, and the use of generative adversarial networks for producing synthetic IOT timeseries data. All of these domains are related to areas that we see emerging or existing applications or represent the foremost domains in AI research today. Secondly, we have expanded to welcome 5 MSc students, supervised by two lead data scientists, myself, and Dr Pedro Baiz (Imperial College).
In future, we want to evolve the sandbox to welcome students from beyond Imperial College, as well as welcome industry partners. We would like to invite industry partners to submit ideas and data for areas of research which would have direct application to themselves, but that they do not have the capacity or expertise to service them.
To read more about the past and ongoing projects, read further into our blog and sign up for regular updates, latest publications and research podcasts.