I recently came back from a 3 month internship at MSR Cambridge. This was the first joint Bing / MSR internship whose aim was to deliver contributions to both the Bing’s production side and scientific research. My work focussed on large-scale label aggregation of human judgments for search engine evaluation using graphical models and the Infer.NET technologies. The main bits of research were new  algorithm for efficient label aggregations for sparse datasets using workers communities structures, and a Bayesian model for combining of text sentiment analysis judgments (awarded as the winner of the CrowdScale challenge). A presentation about my internship work was given in Southampton on Tuesday, 12 November.