(THINKER · 04)
The Thinker pillar — the lab.
Where code
meets cognition.
A small lab at the seam between engineering and psychology — between deterministic systems and the messy logic of human decision-making.
It starts with the fundamentals: working machine-learning problems end to end and writing down, plainly, what I understand and what I do not yet. Each notebook below is public, with its results in the README.
(Published — 2 notebooks)
- House prices — advanced regressionA leak-free scikit-learn pipeline compared across three models with 5-fold cross-validation. Gradient boosting won: CV RMSE 0.1271 (log-dollars), Kaggle public score 0.13211.↗ Notebook on GitHub
- Titanic — survival classificationA first end-to-end binary classifier, with the notes on what I did and did not yet understand. Kaggle public score ≈ 0.76 accuracy.↗ Notebook on GitHub
(Planned — not written yet)
- ／01Vector search for semantic resume parsing
- ／02Fine-tuning Llama on Stoic philosophy
- ／03Behavioral economics in payment UX
- ／04Bayesian models for personal forecasting
Topics, not drafts. They appear on the left when there is something to read.
※ There is no mailing list. New notebooks land on GitHub first.