Heesoo Kim Email & Phone Number

South Korea

• 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗔𝗜 𝗥&𝗗 𝗧𝗲𝗮𝗺
• Analyzed 150k audio samples, boosting Bixby’s speaker verification accuracy by 3.5%
• Developed and trained ECAPA and Depthwise CNN models in PyTorch
• Implemented speaker data augmentation, reducing Equal Error Rate(EER) from 11% to 8.6%
• Applied normalization techniques, further reducing EER to 7.5%

• Created a contrastive learning model to identify misclassified pre-trained architectures, achieving 91% accuracy
• Trained a classifier on 107 unique architectures, attaining 96.55% accuracy
• Fine-tuned RoBERTa model to generate embeddings for a custom dataset

South Korea

• Designed a custom camera app using Samsung Android plugins and Camera2 API
• Collaborated with a team of 5 engineers to integrate the app into Samsung’s Wiki
• Integrated real-time image capture with backend storage, ensuring smooth data handling

West Lafayette, Indiana, United States

• Fine-tuned a Detectron2 model on a custom COCO dataset, accomplishing over 85% accuracy
• Constructed an image annotation tool, enhancing performance by 1.5timesfasterthan the widely used tool,Labelme
• Automated camera calibration, reducing process time from 5 to 2minutes
• Utilized WSL and SSH for efficient development and code updates

Paju, Gyeonggi, South Korea

• Developed a MATLAB script to optimize mortar accuracy, achieving a 5% improvement
• Directed a team of 5 in training and missions, enhancing mortar fire direction

Master Of Science - Ms, Computer Engineering

Graduate Teaching Assistant - ENGR 13300

Bachelor'S Degree, Computer Engineering, 3.58

Two years of leave of absence due to obligatory military service