Built an iOS application from scratch with Amazon AWS supporting for searching the nearest markets with wanted foods available.
>>>>Used AWS EC2, Elastic Beanstalk, S3, Dynamo DB, IAM, STS, etc.
From scratch, built a new Recurrent Neural Network (feature learning based) for Part of Speech task, which can be seen as a RNN base HMM model. Beat the normal HMM and almost beat the HMM with many human designed features.
>>>>I designed the model and implement the main part of RNN.
Under supervisor John Longford (Microsoft Research), use Hadoop to parallel the RGF algorithm. By reading and modifying the RGF code (16000 lines), we finally get 4 times speed-up with 12 Hadoop nodes.
>>>>John commented it as difficulties one in all final projects he proposed, and we finished it well..