We started RobustLinks in 2010 on a National Science Foundation SBIR grant. We pivoted from Prediction Markets to work on Natural Language Processing (NLP) when we realized Social Media content was breaking standard NLP pipelines, which were almost exclusively trained on structured and well behaved corpi (such as Wall St Journal or Thomson Reuters content). Our work focused on how to "embed" words in higher dimensional spaces using Wikipedia as an embedding resource. In 2013 Tomas Mikolov showed how this task can be solved in a neural network, and with that he blew open the beginning of the NLP revolution (ELMO, Transformer, GPT,..).
We continue to stand on shoulders of giants and build on their innovations with work spanning NLP, Computer Vision and more recently transfer learning for time series problems in Finance and Healthcare.