Experience & Skills

Engineered a transformer-based multi-label classification model on LG.com user behavior data, analyzing intent for precise audience targeting for return visits. Achieved a 10% increase in website revisit rate.

Built an NLP pipeline that summarized 50K+ product reviews, using BERT, UMAP, HDBSCAN, and TF-IDF to extract high-signal product features and improve UX quality, leading to a 14% increase in user engagement.

Led psychological segmentation by mapping attitudinal survey data to 20M-customer CDP and deploying tailored marketing content; powered personalized campaigns that achieved a 2x lift in both CTR and CVR through email.

Streamlined call center operation by transcribing audio files using speech-to-text and building a fine-tuned LLM model to extract caller intent and root cause analysis, significantly improving resolution efficiency and agent response time.

Developed a Marketing Mix Modeling (MMM) tool by using Google's lightweight Bayesian framework to evaluate media channel effectiveness and optimize budget allocation, data-driven decisions that improved ROI by 15%.

Optimized factory production processes by solving a convex optimization problem using cvxpy and Gurobi, balancing scheduling, raw material availability, and sales forecasts, resulting in a 15% improvement in factory throughput.

Built a scalable time-series clustering algorithm using Dynamic Time Warping and ANN to group products by forecasted sales behavior, reducing forecast error by 22%, which improved the speed of inventory planning across key product lines.

Specialization in Deep Learning and Big Data. Designed convolutional neural network, computer vision model for classifying people's age, gender, and race. Partnered with MotionFlow, a smart AIoT ad company, delivering inference under 0.5 seconds with over 92% accuracy.

Reduced daily backorder rates to under 2% by automating demand forecasting operation using VBA incorporating Holt-Winters. Designed delivery-date driven prioritization and improving forecast accuracy and speed in demand planning.