Stock Portfolio Tracker (CS50P Final Project)
A small Python app for tracking a stock portfolio with real-time prices and per-user accounts.
Senior BI Analyst · Data Science
Nguyen Manh Tien
I turn banking questions into models and decisions.
Analytics professional in retail banking (unsecured lending, credit cards, acquisition), studying for an MSc in Data Science. I write about credit risk modelling, statistics and building ML systems that people actually use.
- Approval rate
- –
- Bad rate (approved)
- –
- Bads declined
- –
- LightGBM
- XGBoost
- Random Forest
- Probability calibration
- K-Means
- PCA
- Hypothesis testing
- A/B test design
- Power analysis
- SciPy
- Pingouin
- Python
- SQL
- PySpark
- Databricks
- Schema design
- Power BI (DAX, semantic models)
- Matplotlib
- Seaborn
- Unsecured lending
- Credit cards
- Acquisition funnels
- Telesales prioritisation
Model card
about meI came to data from economics, and today I build models and analytics for retail lending at Techcombank: who applies, who gets approved, who is worth a call, who stays, and which levers move those numbers.
My day-to-day sits between business questions and statistics: LightGBM propensity models with calibrated probabilities, lead-generation pipelines on Databricks, and the monitoring that keeps them honest. I'm most interested in the step from a score to a decision: cut-offs, trade-offs, and what each choice costs.
I'm doing a part-time MSc in Data Science and use this site to write up what I learn: credit risk modelling, probability, and building ML systems that people actually use.
- Model
- Nguyen Manh Tien · build 2026.10
- Type
- BI analyst, retraining as a data scientist
- Domain
- Retail lending: unsecured loans, credit cards, acquisition
- Training data
- Economics (FTU) → lending analytics at Techcombank → MSc Data Science
- Languages
- English (IELTS 8.0) · Chinese (HSK 5) · Vietnamese
- Objective
- Turn scores into decisions people can defend
- Known limitations
- Still working through measure-theoretic probability
model_card.yaml
m01Since 2024Lending analytics & modelling at Techcombank
m02Calibrated MLLightGBM propensity & connection models
m032 awardsPeople's Choice · Star of June (D&A)
m04MSc DSVNU University of Science (in progress)
Training log
experienceBusiness Intelligence Contractor
Techcombank
Built BI reports and dashboards for lending portfolio performance before converting to a full-time analyst role.
- Power BI
- DAX
- SQL
Business Intelligence Analyst
Techcombank
Took ownership of the approval-propensity and telesales-prioritisation models alongside funnel, churn and approval-rate analysis for lending products; promoted to Senior.
- Python
- Scikit-learn
- K-Means
- SQL
Senior Business Intelligence Analyst
Techcombank
Build propensity and connection models (LightGBM, calibrated with Platt / isotonic) that prioritise telesales outreach, plus lead-generation pipelines and performance monitoring for credit cards, overdrafts and personal loans.
- LightGBM
- Calibration
- PySpark
- Databricks
- SQL
- Power BI
Validation set
projects
DataCamp Data Analyst Certification: Sales Strategy Review
Evaluating three sales strategies for a new stationery line; analysis in Python and a Power BI dashboard.
A/B Testing for Sales Conversion
Comparing campaign strategies on CR, CTR and CPC with power analysis, bootstrapping and Mann-Whitney U tests.
Notes
writing- 001The Power to Will - Nietzsche's Philosophy Unveiled
Some thought of myself about the will to power of Nietzsche
// End of report