My research focus on Understanding LLMs

Help 20+ students learn Data Science courses in a better way based on my experience and best practices. The data science courses last 6 months, organized in the form of 1 mentor supporting 1 mentee to build skills related to Python, SQL, pandas, data visualization and practical data analysis projects.

My work focuses on developing the Data Science portion of Wondaris (a Customer Data Platform product)● Identity Resolution. Design and implement solutions for incremental and batch matching CDP, which uses a deterministic matching approach.● End-to-end Machine Learning. Leverage Google Cloud to build end-to-end solutions for Machine Learning problems (from processing data, suggesting features to training, evaluating, predicting and monitoring ML models), especially with marketing problems: propensity model, lifetime value prediction, customer segmentation... based on unified CDP data.● Automate feature suggestion. Build a solution that automatically suggests the best features based on clients' input configuration. In addition to meeting many different types of models (classification, regression), controlling and processing input data, this engine must also ensure that overfitting is automatically resolved.● Others. Support customers to debug, suggest improvements as well as participate in solution design for new features related to data science (e.g. RFM).

My work focused on building data warehouses and solving data science problems of VinHomes (a subsidiary of Vingroup).● ETL data. Ingest and cleanse data from different sources of VinHomes then build corresponding data warehouse● Reporting. Build analytical reports about customer segmentation based on purchasing behaviors/objectives and data quality reports, which then are used to propose insights and detailed actions to improve data quality.● Algorithm development. Developed entity matching rules and aligned with business users to consolidate each dimensional dataset, especially customer profile data.● Lead Scoring. Built a Machine Learning model with over 95% precision to automatically classify potential customers (leads) into groups based on online behavior on the website, responses to the agent as well as transactions and interactions with Vinhomes. This model helps increase monthly hot lead conversion rate by 20%.

My work focused on building Machine Learning models to improve the operation of products on ViettelPay ecosystem● Lead Generation. Built a Machine Learning model that follows a look‑alike scenario to find potential users for new products on ViettelPay. The model delivered 2 times higher average conversion rate by messaging than traditional methods for products from a variety of industries.● Credit Scoring. Built a credit scoring model from raw telecommunications data, achieving Gini efficiency greater than 0.7.

Teaching a basic data science course that introduces the process of performing a data science problem as well as modeling algorithms (logistic regression, svm, tree-based modeling, ensemble models)

1. Sentiment Analysis. Built and developed a sentiment classification model that works well with a lot of data from newspaper information.2. Object Detection, Object Recognition. Surveyed AI solutions which are best fit with fore container number (both vertical and horizontal texts).

Study on preliminary design of satellite telecommunications system

Bachelor of Electronics and Communications Engineering*Program: Talented Engineer Training*Rank: Very Good