Welcome to my portfolio! I am an aspiring Financial Analyst with a strong focus on FP&A and data-driven decision-making. This space showcases my hands-on projects in financial reporting, forecasting, credit risk modeling, and investment analysis using tools like Power BI, Excel, Python, and SQL. Through these projects, I aim to demonstrate my ability to turn complex financial data into actionable insights that support strategic planning and performance optimization. Thank you for visiting, and I hope you find value in the work I share here.
Chào mừng bạn đến với portfolio của tôi! Tôi là một ứng viên đang theo đuổi sự nghiệp trong lĩnh vực Phân tích Tài chính (FP&A), với định hướng sử dụng dữ liệu để hỗ trợ ra quyết định chiến lược. Trang này tổng hợp các dự án thực tế mà tôi đã thực hiện, bao gồm báo cáo tài chính, dự báo doanh thu, mô hình hóa rủi ro tín dụng và phân tích đầu tư – sử dụng các công cụ như Power BI, Excel, Python và SQL. Qua những dự án này, tôi mong muốn thể hiện khả năng chuyển hóa dữ liệu tài chính phức tạp thành những phân tích hữu ích, phục vụ tối ưu hóa hiệu suất doanh nghiệp. Cảm ơn bạn đã ghé thăm và hy vọng bạn sẽ tìm thấy giá trị trong những nội dung tôi chia sẻ.

---

Group 1: FP&A

My latest FP&A report: How I built a Dynamic Budgeting Model to optimize Education Operations.

I recently developed a comprehensive Financial Planning & Analysis (FP&A) model for an Education Center to support their FY2025 Budgeting. The goal was to move from static spreadsheets to a dynamic decision-making tool.

Below are key features and strategic insights:
▶ Scenario Planning (Best/Base/Worst Case): Incorporated a dynamic switch in the Assumptions Sheet to stress-test the P&L. This allows stakeholders to instantly see how changes in “Revenue Realization Factor” or “Cash Conversion Rate” impact the bottom line.
▶ Operational Deep-Dive: Instead of a flat revenue projection, I built a Student Flow Model (Beginning + New – Churn). This connects financial outcomes directly to operational metrics like Retention Rate and Marketing Efficiency.
▶ Unit Economics & Capacity Optimization: I analyzed Gross Profit per Class and Teacher Efficiency.
Insight: Calculated the “Financial Breakeven Class Size” to help the Ops team decide when to open or merge classes.
Result: Identified Tier 3 classes (low profitability) for potential resource reallocation.
▶ Marketing Efficiency Analysis: Tracked enrollment conversion by channel (Online Ads vs. Referrals) to optimize the Customer Acquisition Cost (CAC) budget.
This project represents my approach to Finance: Data-driven, Operationally-grounded, and Future-focused.

Group 2: Financial Analysis & Reporting

Project 1: HSBC Financial KPI Dashboard

The HSBC Financial KPI Dashboard project combined Python to extract data from Yahoo Finance and processed to create financial metrics has helped me develop a range of important skills as follows:
• Financial Data Extraction Skills: Using Python (yfinance or other APIs) to collect reliable financial data and automate scheduled data retrieval.
• Data Preprocessing Skills: Cleaning, filtering, transforming, and normalizing raw data to prepare for analysis and visualization. Knowing how to process periodic data, classify, and create data tables suitable for KPI analysis.
• Financial Metrics Calculation Skills: Creating complex metrics from raw data such as Working Capital, Debt Ratios, ROE, ROA, and liquidity ratios. Understanding the meaning of financial metrics and applying them accurately in analysis.
• Advanced Financial Analysis Skills: Evaluating performance trends, financial position, and risks through aggregated indicators. Critical thinking skills to interpret data into valuable insights for management.
• Data Visualization Skills: Designing professional and understandable dashboards with Power BI or similar tools that effectively communicate information. Combining charts, summary tables, and key metrics for comprehensive and visual reporting.
• Project Management & Cross-Functional Collaboration Skills: Planning, organizing, and executing steps from data gathering, processing to complete dashboard design.
Knowledge of banking finance combined with appropriate technology.
• Automation & Optimization Skills: Automating data workflows to save time and reduce errors. Improving processes for periodic or real-time dashboard updates.

Please click here to visit and intereact with the dashboard.

Project 2: Financial Reporting

Tools: Power BI
Focus:
• Asset Management: Monitoring and optimizing asset utilization to ensure financial stability.
• Liquidity Control: Implementing effective measures to maintain adequate cash flow and meet obligations.
• Revenue Analysis: Diversifying income streams across product lines and regions to mitigate risks.
• Credit Management: Utilizing aged trial balance for tracking and minimizing overdue accounts.
Outcomes: Leveraging data visualization to support strategic decision-making and long-term planning.

Please click here to visit and intereact with the dashboard.

---

Group 3: Sales Analysis and Forecast

Project 1: Building a Robust Automated Valuation Model (AVM): A Stacking Ensemble Approach

In Real Estate Finance, precision is profit. A slight deviation in asset valuation can lead to significant portfolio risk. I recently engineered a Housing Price Prediction model to challenge the accuracy of traditional linear appraisals. By treating the Ames Housing dataset as a financial modeling problem, I aimed to minimize valuation variance.

The Financial/Business Logic:

• Asset Quality vs. Size: My analysis confirmed that Overall Material Quality (OverallQual) is a stronger predictor of value than raw square footage. This implies that “renovation premium” is a quantifiable metric in asset pricing.
• Market Non-Linearity: Real estate prices do not follow a straight line. I applied Log-transformations to the target variable (SalePrice) to normalize skewness, ensuring the model handles high-value outliers (luxury properties) effectively.

The Data Science Engine:

• Data Integrity (MICE): Instead of dropping missing financial data, I used MICE (Multivariate Imputation by Chained Equations). This preserves the statistical correlations between features—crucial for maintaining data fidelity.
• Stacking Architecture: To reduce model bias, I implemented a Stacking Regressor combining Ridge, Lasso, ElasticNet, Gradient Boosting, and LightGBM, with XGBoost as the meta-learner.

Results: RMSE: 0.1158 – effectively minimizing the spread between predicted and actual market values.

This project reinforces my belief that advanced ML ensembles are the future of Real Estate Appraisal and Risk Assessment.

Project 2: Pharmacy Sales Report and Cross-Selling Analysis

Tools: Python + Power BI

Through the actual project, including analyzing revenue’s attributes and proposing cross-selling, below techniques are what I could learn:

Cross-selling Analysis with Python
• Handling and preprocessing large real-world sales datasets using pandas, including detecting and filling missing values, standardizing dates, and continuously checking time series data to avoid missing entries.
• Implementing cross-selling analysis: identifying the main product consumed by each disease group, automatically generating probable cross-sell products along with ranking by subgroup revenue, and generating practical cross-sell recommendation tables for each disease group.

Sales Analysis and Power BI Visualization
• Designing multi-dimensional Power BI dashboards, connecting multiple tables such as revenue, SKU groups, suppliers, products, customer groups, and disease groups—all dynamically analyzed using table filters, time slicers, and hierarchical drilldowns.
• Implementing multi-table data models, DAX calculations for crucial KPIs (sales, achievement ratios, number of days above target, growth rate, sales/order), and conducting comparative analysis by month/year/week.
• Building visuals and dynamic tables, harmoniously combining cards, summary tables, time series plots, and waterfall charts to communicate financial data very clearly to senior management.

Project 3: Business Forecasting by Excel

Tools: Excel
Focus:
• Time Series Analysis: Applying methods like Naive, Moving Average, and Exponential Smoothing.
• Regression Models: Leveraging statistical relationships for forecasting outcomes.
• Judgmental Forecasting: Integrating opinions to enhance forecast accuracy.
• Model Selection: Emphasizing the importance of aligning forecasting models with data characteristics.
Outcome: Improved forecast accuracy and model selection

Click here to reach the full report.

---

Group 4: Investment & Portfolio Analysis

Projetc 1: Financial Portfolios, Modern Portfolio Theory & Asset Pricing by Python

Tools: Python (Numpy, Pandas, Yfinance, Matplotlib, Seaborn)
Focus:
• Data Preparation: Cleaning, handling missing values, and transforming data for analysis.
• Data Visualization: Using scatter plots and bar charts to illustrate trends.
• Portfolio Optimization: Calculating Sharpe Ratios and the efficient frontier for selecting optimal portfolios.
• Regression Analysis: Predicting future stock prices and returns for informed financial decisions.

---

Group 5: Credit Risk Modeling & Management

Projetc 1: Bridging Data Science & Credit Risk: From Raw Transactions to Strategic Policy

Bridging Data Science & Credit Risk: From Raw Transactions to Strategic Policy
Data Science in Finance isn’t just about model accuracy—it’s about minimizing financial loss.
I recently completed an end-to-end analysis of Credit Cardholder behavior, using Python to solve two core banking challenges: Customer Segmentation and High-Risk Prediction.

Here is how I translated technical algorithms into business value:
1. Unsupervised Learning (The “Market Strategy” Angle)
Instead of arbitrary grouping, I used K-Means Clustering to identify distinct customer archetypes.
– Technical Decision: I prioritized K-Means over density-based methods (DBSCAN) to ensure complete customer coverage for operational feasibility.
– Business Outcome: Identified 4 actionable segments, including a critical “Cash Advance Focus” group—customers using credit as short-term liquidity, signaling high default risk.

2. Supervised Learning (The “Risk Management” Angle)
I built a predictive model to flag high-risk users (potential defaulters).
– The Challenge: Imbalanced data (defaulters are rare).
– The Solution: I deployed XGBoost with weighted scaling, outperforming Logistic Regression and SVM.

The Metric that Matters: I optimized for Recall (Sensitivity) rather than just Precision. In credit risk, a False Negative (missing a defaulter) costs significantly more than a False Positive (auditing a safe customer). My model achieved 85.4% Recall.

Key Financial Insight:
My feature importance analysis revealed that Tenure is irrelevant to risk. Instead, Payment Discipline (Percentage Full Payment) and Liquidity Stress (Cash Advance Frequency) drive >80% of the risk profile.
This suggests banks should shift approval criteria from “Spending Volume” to “Payment Ratios” to reduce delinquency exposure.

Read the full Technical Report attached for the code logic, visualizations, and detailed policy recommendations.

Project 2: Credit Risk Modeling

Tool: Python
Focus:
• Credit Risk Fundamentals: Explaining key concepts like Probability of Default, Loss Given Default, and Exposure at Default.
• Feature Selection: Highlighting methods to identify relevant predictors for credit risk models.
• Model Implementation: Demonstrating logistic regression for predicting the Probability of Default.