NLP at The Machine Learning Center of Excellence (MLCOE)

- Worked with the Video Understanding for Relevance team on the research project of multimodal video sentiment and emotion recognition

- Worked with Columbia Speech Lab, led by Prof. Julia Hirschberg- Projects: Humor Detection in Multimedia Content, Sentiment Analysis on Low-Resource Languages (DARPA LORELEI Program), Speech Emotion Conversion- Work published in Proceedings of 3rd IEEE Conference on Multimedia Information Processing and Retrieval (IEEE MIPR’19)- Paper submitted to special issue on Affect Recognition in Multimodal Language, Cognitive Computation Journal

- Worked at Waterloo AI Lab, led by Prof. Pascal Poupart- Designed a model framework for a chatbot that integrates the use of chat history into generative sequence models(RNN). - Implemented models (K-nearest, linear regression) for selecting most relevant chat history entries.

- Worked on FrostEd, the Frostbite Integrated Development Environment. Developed new features for connecting to external text editors through FrostEd, and synchronizing database and user interface in runtime with externally edited data.- Implemented a language server for GraphScript, a scripting language developed by EA, which supports auto code completion and debugger for GraphScript in Visual Studio Code through runtime communication with FrostEd. - Built a Visual Studio Code extension for GraphScript to support syntax highlighting and code snippets.

- Worked with Professor Pascal Poupart on Emofitness, a project aiming to analyze people’s anxiety levels and emotional fluctuation, and to predict possible stress and depression in individuals.- Conducted thorough data analysis with plausible accuracy, applying Hidden Markov Model, artificial neural network, and Support Vector Machines on detrended time series data. Validated results using k-fold cross-validation and confusion matrix.

- Worked on Office Lens Android application. Designed and implemented new features from scratch for taking and uploading multiple photos. New release ready to ship.- Enabled and enhanced Office Lens iOS bidirectional language support.- Retrieved information from raw data and created user action telemetry reports fordebugging and project management purposes.

- Built up automation testing based on .NET framework for testing endpointson servers.- Implemented automation programs to generate massive data for demonstratingand testing purposes.- Created automation tools for processing SQL scripts and database schemasto generate useful results.

- Monitored high priority SP automated testing on SAP Sybase SQL Anywhere family of products.- Built day-to-day automated testing for pre-released SQL Anywhere builds. - Diagnosed testing errors by parsing server log files and editing test scripts. Reported performance failures when necessary.- Performed automated and manual testing on pre-released software builds against all supported consolidated database servers such as DB2, Oracle and MySQL. •Configured hardware and software environment on Unix virtual machines and devices for testing.