Dr Andrey Kravchenko

My DPhil qualification has induced me to explore the intersection of machine learning and unstructured data extraction. I had a significant role in the DIADEM project, which produced state-of-the art research in the field of large-scale fully automated web data extraction. The technology underlying DIADEM was later commercialised as an Oxford University spin-off Wrapidity, which was acquired by Meltwater, one of the most predominant world-wide media intelligence companies. My current research focus is on the theory and application of anomaly detection in big data using sequences and graphs, and in particular, the development of efficient machine learning algorithms based on the embedding of vectors. I also work on exploring the broader connection between black-box machine learning models and knowledge-based systems, with a particular focus on knowledge graphs. Machine learning models do not involve human expertise and are often difficult, if not impossible, to explain. On the contrary, knowledge-based systems are easy to translate into an interpretable form but are based on human domain expertise and therefore require significant manual work to build. I would like to work towards bridging the gap between these two different aspects of artificial intelligence in my research. I have a strong interest in scientific collaborations. If you have a project which may relate to my area of expertise, please contact me to discuss ideas of mutual interest and potential collaboration.

My research has always had a strong focus on real-world applications, as well as potential business opportunities. I am actively involved with DeepReason.ai, an Oxford University spin-off, which delivers large-scale knowledge graphs to clients across a wide range of industries based on a unique platform with a novel reasoning language created as a result of research done here at Oxford. I have also founded a web and mobile app development agency, Vitruvian. We develop a wide range of software products with a particular focus on products that use algorithmic and state-of-the-art deep learning models, such as transformers and stable diffusion.