Ting-Yu Liu

Designed and built an AI-based object detection pipeline for Carrier drilling inspection, automating visual checks and reducing inspection time and manpower by 30%. Developed an AI-driven OCR pipeline for product documents and nameplate images, integrating outputs into a NoSQL database to automate data entry and cut processing time from 25 minutes to 3 seconds with 93% accuracy. Trained ML models (Regression and Decision Tree) with process engineering teams to identify optimal manufacturing conditions for Carrier yield improvement. Maintained and fine-tuned AI defect detection models across 2 sites with retraining based on performance metrics and production feedback. Built a real-time, interactive dashboard for data visibility across 10 sites in 2 countries.

Developed and deployed an ML-based churn prediction model (Decision Tree) to enable proactive retention, reducing churn by 3%. Built a deep learning-based competitor detection model on Linux with automated data pipelines, reducing unwanted calls by 40% in one month. Designed and deployed an end-to-end ML system on Linux for ETL scheduling anomaly detection, from model training to real-time dashboard deployment, reducing maintenance time by 30%. Analyzed customer behavior to segment customers and improve targeting and advertising efficiency.

Collaborated with SAS engineers to build ETL pipelines for O’bank, supporting credit risk modeling data. Developed customized Customer Due Diligence (CDD) modules within SAS AML systems for multiple banks, integrating SAS with MS SQL to identify high-risk customers.