• Improved the core model accuracy by 6% (highest in company history) by implementing a noise-robust dual graph neural network framework on drug discovery tasks. • Expanded the training data by 12% by developing a novel label-generation method further increasing accuracy by 5%, leading to more reliable predictions of bio-molecular interactions. • Utilized TensorFlow’s mirrored strategy and distributed training across 4 GPUs for efficient parallel processing. - 약물 탐색 작업에서 노이즈에 강한 듀얼 그래프 신경망 프레임워크를 구현하여 모델 정확도를 4.6% 향상 (회사 내 기록 최고치 달성). - 새로운 레이블 생성 방법을 개발하여 훈련 데이터를 12% 확장, 모델 정확도 5% 추가 개선 및 생체 분자 상호작용 예측의 신뢰도 향상. - TensorFlow의 미러드 전략 및 분산 학습을 활용해 4개의 GPU에서 효율적인 병렬 처리 수행.

더보기

• Accelerated customer onboarding time by 400% by developing an in-house recommendation algorithm, automating the consultation pipeline. • Defined research and project objectives and database specifications supervising a team of 4, reporting to a higher manager and the COO. Played a pivotal role in service development, led research meetings. • Decreased the data error rate by 14.4% by implementing and overseeing the company-wide transition to a semi-automated database management and visualization system. - 고객 온보딩 시간을 400% 단축하는 추천 알고리즘을 자체 개발하여 컨설팅 파이프라인을 자동화. - 팀원 4명을 지도하며 연구 및 프로젝트 목표 설정, 데이터베이스 명세 정의 및 COO 보고 업무 수행. 서비스 개발과 연구 회의를 주도. - 회사 전체에 반자동 데이터베이스 관리 및 시각화 시스템 전환을 이끌어 데이터 오류율을 14.4% 감소. - 데이터 품질 보증 프로그램을 개발해 2,000개 이상의 오류를 수정했으며, 크롤링 스크립트를 활용해 데이터베이스를 4배 확장.

더보기

>> Processing and Analyzing recorded bio-signals to identify key health attributes of the subject using MATLAB. >> Recording live bio-signal (ECG, EEG, EMG, PPG) from individuals preventing artifacts and reducing noise using various bandpass filters.

더보기

>> Assisting in Ph.D. research by cleaning and visualizing fMRI data using Python and MATLAB.

더보기

Full-stack development of university-related applications for course registration, forum discussions, event announcements, and transportation arrangements. >> Developing an ecosystem of five (currently) Python+React progressive web applications supporting KAIST's academic management and authentication systems. >> Maintaining the above services with 1000+ users daily by optimizing CRUD operations.

더보기

>> Currently working on CAR T Cell production. >> Dextran-based Chemical Hydrogel production (full process). >> Cultivated Human Embryonic Kidney (HEK293T) cells with 95% confluency and 92% vitality 3 times in a row. >> Successfully transfected HEK293T cells with Green Fluorescent Protein as measured by a confocal microscope. >> Performed essential laboratory techniques: cell incubation, centrifuging, aspiration, cell dyeing, extrusion...

더보기

>> Developed a GCP (Google Cloud Products) progressive web application for verification and cross-referencing of micro-loan applicants. >> Implemented authentication and transaction information fetch functions using Open Banking API in Python, TypeScript. >> Reduced storage load by 14.2% by redesigning the database schema and eliminating unnecessary cost.

더보기