Welcome to Whizzlab, our data and AI lab in the cloud :).

We work with health data and clinical artificial intelligence to answer problems with practical, real-world applications.

Recently, we’ve been: using natural language processing models to automatically characterise and monitor the entirety of clinical AI research; building a national database of critically ill COVID-19 patient referrals to support pandemic efforts, research into treatment effectiveness, and machine-learning based phenotyping; using graphs (the network type, not the pie chart type) to analyse national data flow networks; and bringing down practical barriers to getting artificial intelligence deployed.

See our Projects page for more details. If these things sound interesting, or if you have some ideas you want to explore, contact us to collaborate! We love to meet new people: hello@whizzlab.ai (Joe + Steve).

Recent Posts

• NHS England cloud infrastructure through freedom-of-information requests
  The recent (and at time of writing, still on-going) catastrophic failure of IT systems at a major NHS Trust has put ageing NHS IT infrastructure into the spotlight. The extent of data loss is yet unclear, but the failure of both primary and the single off-location back-up unit has clearly prevented disaster recovery plans from… Read more: NHS England cloud infrastructure through freedom-of-information requests
• Data-driven insights during a pandemic
  How do you approach a pandemic disease of which very little is known, where resources are scarce, and where treatment decisions become highly consequential? Data that is high quality and labelled with high confidence can be used to model complex associations, and models can be used to predict future events (supervised learning). But without these… Read more: Data-driven insights during a pandemic
• A framework for development maturity in clinical artificial intelligence
  Out of the large amounts of published clinical AI work, it’s important to distinguish between research that focuses on improving model accuracy on datasets, to research that is undertaking some sort of process to evaluate AI in real-world conditions. This includes comparative research, which assess performance of a model against a real-world alternative (and from… Read more: A framework for development maturity in clinical artificial intelligence
• Bringing biological ARDS phenotypes to the bedside with machine-learning-based classifiers
  This manuscript was published in The Lancet Respiratory Medicine in January 2022 (https://doi.org/10.1016/S2213-2600(21)00492-6), and addresses key problems with current attitudes to machine learning on electronic health record data. It’s not enough to train a good model on a pre-curated dataset from a research setting. The work that’s needed is in data and EHR infrastructure, to… Read more: Bringing biological ARDS phenotypes to the bedside with machine-learning-based classifiers
• The “promise” of AI in critical care – talking about the wrong things
  The volume of complex, high-temporal-resolution data that intensive care physicians encounter every day is hard to interpret, and it is certain that many useful signals for predicting or prognosticating exist that are entirely unknown. Intuitively, it is the perfect situation for AI algorithms to make an impact, but recent reviews of AI in critical care… Read more: The “promise” of AI in critical care – talking about the wrong things