Junhong Kim

- Improve performance: Identified performance degradation from the pipeline, suggested new approaches, and implemented the update with thorough tests bringing a 20% performance increase. - Collaboration: Guided new employees and colleagues from Java backgrounds to help them pick up Python and Test Driven Development (TDD) by running a weekly session. - Preprocessing: Implement random sampling from Mongo and Nutch Elastic Search databases to replicate current data distribution using Python.

- Improvement: Redesigned internal human face detection algorithm for video to make it robust in various environments and improved 3000% detection rate for corner cases. - Deeplearning: Implemented, trained, analyzed, and deployed multi-modal (images, audios, and texts) deep learning model using Pytorch and Tensorflow utilizing Transfer-Learning, Semi-Supervised Learning, Attention, and XGBoost improving model performance by 10%.

- VBA Parser: Implemented VBA code parser on an enterprise actuarial software using regular expressions and ANTLR, enabling auto-complete feature for writing VBA scripts and go-to-definition feature. - Parallel Processing: Implemented a background worker queue, improving usability and loading speed up to 20%, utilizing BackgroundWorker and Queue. - Maintainability: Drove refactoring of entire code base to MVC pattern and refactored all internal source code, significantly improving testability.

- Research Assistant - Human Computer Interaction: Implementing data collection application in Android and designing deep learning model with Keras to tell the voice is from adult or child. - Research Assistant - Support Experiments: Creating simulated situations with C and designing a survey.