About
I'm Tung Anh Tran, a data scientist building machine learning models for classification and computer vision on large, complex datasets. Most of my work is in healthcare and education, and I'm open to problems in other industries.
Background
Placeholder
Two or three sentences on where I trained, the teams I have worked in and the kind of data I have spent most time with.
One or two sentences on what I am looking for next, and the problems I would most like to work on.
How I work
Calibrate out of sample
A calibrator scored on its own training predictions looks better than it is. In the DaT challenge, the one that looked best in-sample did worse than no calibration once it was cross-validated.
Break results down by group
A pooled score can hide the groups that matter most. The same baseline looked well calibrated on all scans and was far off on its smallest acquisition family.
Stack models that fail differently
A weaker model that is wrong on different cases can add more than a stronger copy of what is already there. The tree model, weaker alone, carried a third of the final stack.
Toolkit
Methods
- 3D medical image preprocessing
- Convolutional neural networks
- Probability calibration
- Cross-validation design
- Stacking
- Feature engineering
- Nested hyperparameter tuning
- Test-time augmentation
- Paired significance testing
- Code-execution submission packaging
- Data-compliance guardrails
- Wage equations
- Administrative panel data
- Official statistics writing
- Event data modelling
- Data validation at the source
- Feature engineering with pandas
- Privacy-safe data publishing
- Synthetic data generation
- Baseline forecasting
- Dashboard design
- Unit and app testing
- Continuous integration
Tools
- Python
- PyTorch
- LightGBM
- scikit-learn
- NumPy
- pandas
- SciPy
- pytest
- Docker
- Claude Code
- uv
- SQLite
- Streamlit
- Plotly
- ruff
- GitHub Actions
In brief
- Based in
- City, country Placeholder
- Vietnamese
- Native
- English
- Fluent (TOEIC 965/990 – C1 level, expired 2024)
- French
- Professional working proficiency
- Experience and education
- See Work