• Built "Gas Repurchase Behavior Model" to forecast 200K existing customers' repurchase time and new customers' time, aiming to help predict the daily demand for gas tank and improve capacity for gas logistics. Implemented LSTM Model with PyTorch to estimate aggregate demand and ensembled with Curve Fitting Model for Repurchase Time with SciPy. Achieved a 0.85 precision rate for the most valuable regular-repurchase customer with 20% MAE. • Built, Deployed, and Iterated "Revenue Measurement" from scratch with the financial team to support logistic decision and budgeting decisions. Analyzed the Monthly Timeseries of sale of all-products and seasonal behavior of gas demand in different areas using Timeseries analysis model like Prophet and LSTM model with seasonal patterns. The MPE is less than 5.0% for three consecutive recent months.• Full stacked "Inventory Detection and Classification Model" using YOLOv8 and Segment Anything Model. Organized a Labeler to create the dataset of tank by interactive web and speeded up collection by >100%. Exploited SAM for instance segmentation for uploaded inventory picture with mAP of 97.8% and Fine-tuned YOLOv8 model with customized dataset for classifying used/unused tank with f1-score of 90.1%. Deployed for mobile app using FastAPI interface for real-time recognition.

• Built "Parking Space Usage Prediction Model" for over 6K treasury customers of all eight cities in Taiwan, with MAE 11.9%, causing ROA increase 13% in the first season of deployment by eliminating less profitable parking spaces.• Designed algorithms and models to find the efficiency of each delivery route for egg collecting truck. Implemented multiple models, such as VRPTW Model, to facilitate the analysis of the Egg Selection-Egg Farm relationship. Maximized the routing model’s effectiveness and rewarded farmer's satisfaction.• Built Routing Tracking Dashboard for Truck Drivers to monitor performances and daily operation cost such as gasoline fee using Power BI connected to third-party GPS tracker.• Automated reporting DataViz pipeline for auditing workflow with Python (openpyxl, xlwings) to reduce 2 hours of manual work daily.

• Auditing tax and financial statements for more than a dozen companies with total revenues of NT$ 200 million and one hundred total works.• Advised key tax and equity strategies for client corporations and small-median enterprises, including issuing preferred stocks, and intra-corporation debts.• Organized and maintained financial forecasting statement for bank loan leveraging.

• Participating in the organization, acquisition, development and management of property’s investments• Engaging in the market research and product-pitching for the development of commercial properties• Building financial models to evaluate the profitability and feasibility of property acquisition projects• Drafting operation proposals, business plans and presentations and for various investors and clients• Structuring and maintaining the database of Taipei office market and Taiwan property transaction