● Collaborated with various teams to design, experiment and deploy ML solutions for the e-commerce platform● Discovered monetary loopholes within the rebate pipeline and prepared labeled datao Conducted feasibility studies that demonstrated a potential US$200,000 annual saving using variousmodels such as xgboost, catboost, and LightGBMo Created interactive Power BI Dashboard to finalize model and thresholdso Designed, tested, and iteratively improved Genrative AI (GPT) prompts to label data to scale up themodel for production● Designed product classifier using DNN to recognize specific Microsoft products for Product Managers● Fine-tuned BERT model to produce a product categorizer to categorize millions of daily products● Created augmented conversion data by connecting the dots from different data sources for domains that didnot opt-in our data collection services● Designed, deployed and maintained various MLOps pipelines which led to better analytics qualities fordownstream teams

● Conducted a data enrichment feasibility study using ML for a large financial institution located in Alberta● Provided two different solutions to find geo-informations for Point of Sales terminalso Created a classifier pipeline that leverages user data to find the geo-informations in conjunctionwith Google places APIo Derived a scoring formula inspired by astrophysics to approximate the geo-information usinggradient descent● Created a pipeline that enables data scientists to link categories to Point of Sales terminalso Integrated previously under-utilized Mastercard data to transactions using Regexo Improved the Mastercard categorization with NLP (Fuzzy Matching)o With the integrated data as a training set, predicted the categories of the unmatched PoS terminalsusing tf-idf and XGBoost● (Internal) Created a dashboard that helped the executives to better understand employee training

● Conducted A/B testing regarding product placement on 1200 Stores located across Beijing and Shanghaio Variation/Control split using an algorithm derived from KNN and validation by Kolmogorov–Smirnov test.o Presented weekly test results and forecasts to data specialists and stakeholders using Power BI Dashboards● Optimized sales via product placement and store profiling using Network Analysis and Hierarchical Clustering● Using Ensemble Voting and AdaBoost, maintained and updated the customer segmentation model● NLP (Sentiment Analysis) of user feedback to gain insights into public opinions of new e-commerce products● Automation of data parsing and cleaning process using Python and SQL

● Data Integration of newly acquired factories, including quality verification, anomaly detection, and pivoting● Validated, maintained, and updated time series predictive models, and created forecast visualizations using Seaborn ● Design and implementation model using XGBoost to recognize carelessly filled surveys