반도체 설비 이상 감지 모델 개발 (삼성전자)

• 반도체 설비 이상 감지를 위한 AI 모델 개발

• 다양한 시계열 센서 데이터를 이미지로 변환

• Unabeled 데이터의 활용을 극대화 하기 위한 최신 준지도학습 방법 제안

• 보지 못한 이상 유형을 탐지하기 위한 OOD Detection 방법 제안

Quantum Computing for Optimization (Sabre)

•  Collaborated with a leading quantum computing company and faculty from multiple universities

•  Converted optimization problems into QUBO (quadratic unconstrained binary optimization) format

•  Shared airline domain knowledge and proposed some decomposition techniques to feed optimization problems into the quantum computer

Malaysian Name Ethnicity Classification (Sabre)

•  Developed an algorithm to identify ethnicities from full names

•  Used deep learning algorithm such as RNN to predict unknown ethnicities

•  Used other machine learning algorithms (e.g., boosting, bagging and random forest) for comparison

Value Scoring for Airline Bundles (Sabre)

•  Developed an algorithm to estimate a true value of airline bundle (e.g., flight + free baggage + lounge access) and a value of individual ancillary

•  Used various machine learning algorithms (e.g., Lasso regression, Xgboost)

Hotel Channel Optimization (Sabre)

•  Developed an algorithm to allocate the optimal number of rooms for each channel (e.g., direct booking, online travel agency)

•  Used various machine learning algorithms to forecast upcoming demand

Appointment Scheduling at a Multidisciplinary Outpatient Clinic (Dell Medical School at the University of Texas at Austin)

• Investigated the problem of constructing an appointment template for scheduling patients

•  Developed two-stage stochastic program to account for the randomness and the sequential nature of the decision

•  Proposed valid inequalities, limiting indices, and the use of special ordered sets helped to speed up the computations

Hierarchy Machine Setup (Texas Instruments)

• Developed an optimization model for hierarchy machine setup at assembly and test facilities

•  Determined machine setups, lot assignments and lot sequences

•  Implemented a multi stage heuristic algorithm to schedule lots using C++ and COIN OR

Shift Scheduling with Flexible Breaks (European Airport Ground Handlers)

•  Developed a shift scheduling problem with flexible breaks for airport ground handlers

•  Implemented a rolling horizon algorithm for demand fluctuations using Java and CPLEX

•  Used rolling horizon reduced undercoverage by up to 41% compared to scheduling breaks in advance

Optimal Detector Location (BNSF Railway)
*BNSF Railway is the largest freight railroad network in North America

• Implemented a mathematical model to find an optimal location to install detectors for railcars

•  Developed a greedy heuristic algorithm for a large scale problem to find a good solution in a short time

•  Proposed some algorithms to reduced candidate locations by 32% with less than 1% coverage difference

Prepaid Card Business Development Strategy (Accenture)

• Estimated market size, business planning with top level revenue potential and concretized business model with details of revenue scheme