고성운 Email & Phone Number

Seoul, Kr

Seoul, South Korea

Seoul, South Korea

• Created xgboost model for brand item matching initiative. Improved matching accuracy in brand electronic products by 40%.• Designed and implemented async-to-sync service for embedding based candidate retrieval. Improve candidate recall by 20%.• Applied image features for matching. Improved recall by 18% with precision of 98%.• Designed and implemented dataset based matching model test service with UI. Allowed comparison between models.• Applied deep learning FaceNet model for image matching, image based models improved recall by 2%.

Greater Seattle Area

• Designed and implemented a real-time service for predicting promotional effect on demand forecast.• Implemented a map-reduce pipeline to import terra bytes of serialized data.• Provided a structural way to access values in serialized data.• Reduced operational load by changing applications on Network File System to use Distributed File System.

Greater New York City Area

• Researched efficient means to define relationships between videos from social video websites through image indexing techniques and clustering methods.

Greater Seattle Area

• Implemented Iterative Map/Reduce feature to the pipeline for Hadoop Framework by reverse engineering code base.• Parallelized CART (a decision tree construction algorithm) using Iterative Map/Reduce to cluster tens of millions of records.

• Complied technical documents for multiple environmental mark qualification criteria for house hold appliances.

West Lafayette, In

• Implemented an object recognition algorithm for a finite set of objects to determine grasping points of the objects.• Developed a collaborative system of robot arms to improve the performance and to expand the workspace.• Designed a closed loop system for visually servoing the robot arm to reduce the position error of the robotic arm.

Master'S Degree, Computer Science

Bs, Electrical Engineering