October 5, 2012
I work in London, but I travel in from Milton Keynes
Currently I'm using the scientific Python stack (numpy/scipy/matplotlib) for inference on text, including topic-modelling, supervised topic-modelling, correlated topic models and conditional word models. Previously, in my commercial work, I'd experimented with Hadoop and Mahout for regression, and did some work on bandit algorithms for recommendation systems.
I've been working in machine-learning for five years now, and have been interested in data-science for longer. In that time I've worked on record-linkage, text-classification, recommender systems, and now inference on text (topics, word models). It's been very interesting to me to see how machine-learning has taken off, particularly in London in recent years, and I'm curious to see what problems people are tackling and how they're tackling them.
I'm a PhD student currently looking into topic modelling (LDA, CTM) and conditional text models - essentially inference and text. Prior to that I worked on recommenders for Cognitive Match and text-classification & virus-detection ScanSafe/Cisco