We're looking to solve hard problems with innovative solutions, using the latest techniques in the field of machine learning.
We have expertise in using deep architectures for a range of tasks, including image classification, semantic segmentation and unsupervised generative modelling.
Bayesian statistics forms a principled framework for not only getting your data to predict the answer to your questions but also to find out the uncertainty in those predictions.
Utilising powerful yet understandable models allows for the creation of clear visualisations, resulting in a concise explanation of the data that highlights areas for focus.
A quick look at some of the things we're currently working on. We do research and development work both for clients and as in-house projects.
For Cambridge University and Toyota Motor Europe. Using deep convolutional encoder-decoder nets to label pixels in road scenes in order to understand the car's environment.
Using deep density models to learn probabilistic models of high dimensional continuous inputs and then answering inferential queries under arbitrary constraints in these models. For example predicting plausible 3D human poses consistent with the 2D information given by a skeletonisation of a photograph.
Prototype nutrition app, utilising DNN classifiers to recognise the presence of fruit and veg in food images in order to estimate nutritional value.
Kesar has several years experience working on computer vision systems, has an interest in tackling novel data science challenges. He holds an Information Engineering Masters from Cambridge University.
Matt is a PhD student at Edinburgh University's Informatics department. His research is focused on Bayesian statistics and sampling methods and he has practical experience training competitive deep learning systems.
If you have some interesting data you'd like to analyse or want to chat machine learning, please leave us a message!