Data science startup Run:AI announced July 8 that it is providing technology to a leading British medical research center to help it better manage AI resources and provide elastic resource allocation, visibility and control.
The London Medical Imaging & Artificial Intelligence Centre for Value Based Healthcare uses a huge trove of de-identified patient data held by the NHS, including medical images and patient clinical pathway data, to train sophisticated AI learning algorithms. These algorithms are used to create new tools for faster diagnosis, personalized therapies, and more effective screening.
Since the outbreak of the COVID-19 pandemic, the AI Centre has devoted much of its resources to the fight against the novel coronavirus. It recently contributed an AI diagnostic tool that found that the loss of the sense of taste and smell is a stronger predictor of COVID-19 infection than fever, which resulted in the UK government amending its official advice on suspected infections.
Run:AI ensures that the AI Centre's data scientists can get the full use out of their hardware, guaranteeing that graphics processing unit resources are efficiently and elastically allocated to teams that need them. This enables the AI Centre to run more experiments and to speed up time to results, while providing cross-team visibility into how their hardware is being used.
"Healthcare is one of the most important and impactful uses of advanced AI, especially now as it can help save lives during the COVID-19 pandemic. We're proud to be working with the London AI Centre to help ensure their important research can get the best use out of their hardware, so they can run more experiments quickly and efficiently," said Omri Geller, CEO and co-founder of Run:AI.
Since installing Run:AI, the AI Centre has slashed the time taken to complete its experiments. The current average is just a day and a half, whereas a simulation of the AI Centre's exact infrastructure running without Run:AI showed an average of over 46 days per experiment - an improvement of 3000%. Over a 40-day period, the researchers ran more than 300 experiments after installing Run:AI compared to just 162 in a simulation of the same environment over the same time period. In addition, actual GPU utilization doubled in the months since Run:AI's platform has been in use.
"Our experiments can take days or minutes, using a trickle of computing power or a whole cluster," said Dr. M. Jorge Cardoso, Associate Professor & Senior Lecturer in AI at King's College London and CTO of the AI Centre. "With Run:AI we've seen great improvements in speed of experimentation and GPU hardware utilization. Reducing time to results ensures we can ask and answer more critical questions about people's health and lives."