A place to write about data, software, and whatever catches my curiosity — free from the pressure to monetize or to be right. Exploration, for the simple pleasure of it.
Browse: Data·Image Processing·Cryptography·Personal
Dynamic Email Templates - Part 3: Stopping on Regret · latest
How we turned 'is my test done?' into a statistical verdict using an expected-regret stopping rule, and the lifecycle machinery around it: promoting winners and graduating finished tests into static templates.
Dynamic Email Templates - Part 2: Allocating on Confidence
A field guide to allocation strategies for per-slot email bandits: probability matching versus proportional rules, per-slot rewards, cold-start gates, guardrails, and shrinkage - the math, the strengths, and the failure modes of each.
Dynamic Email Templates - Part 1: One Bandit per Slot
How we evolved from A/B testing whole emails to dynamic templates with per-component variant pools, using independent multi-armed bandits per slot to escape combinatorial explosion - covering the architecture, the weights table, and the feedback loop.
Paint By Numbers - Victory At Last?
After a 10-month hiatus, the paint-by-numbers project reaches its conclusion. Discover how switching to CIELab color space, leveraging morphological transformations, and refining the algorithm produced stunning results with organic shapes and natural colors.
Winning with Multi-Armed Bandits: Smarter Experimentation in Databricks
Learn how Multi-Armed Bandits provide a powerful alternative to static A/B testing by balancing exploration and exploitation, using Bayesian methods to adapt experiments in real time and maximize results.
Simple Anomaly Detection in Databricks Using Mean and Standard Deviation
Learn how Consumidor Positivo built a simple yet effective anomaly detection system in Databricks using mean, standard deviation, and generative AI for rapid data quality monitoring, automated diagnosis, and actionable insights.
An Attempt at Paint-by-Numbers - Part 3
Explore advanced techniques for removing thin regions, detecting boundaries, and labeling regions in images to achieve a classic Paint-by-Numbers effect using Python and OpenCV. This post covers practical algorithms, code examples, and visual results for image simplification and region labeling.
An Attempt at Paint-by-Numbers - Part 2
Explore advanced image processing techniques to transform photos into paint-by-number artworks. This post covers morphological transformations, region analysis, and Python code optimizations to simplify complex images for creative projects.
