Code by Night Learning to Program Journey Log

I tested Prophet against seasonal naive and ETS on two monthly series to see why a clean forecast chart is not enough to trust the model.

How to reduce e-commerce churn after building a churn prediction model.

ARIMA is old, limited, and a little annoying.

AI and automation save me time but they also make it much easier to build small programmes that make life very difficult for me later.

I already understood AI agent guardrails in theory, but writing a tiny path check made me respect the annoying details behind them.

After one year of using Omarchy, the thing I value most is not the Linux setup itself.

After one year working around AI security and governance, I trust flashy AI demos less and pay more attention to data, permissions, discovery, and the boring systems around AI.

Churn prediction looks like a clean data science problem, but useful retention work needs systems, ownership, context, and feedback loops.

After finishing my data science and machine learning master’s, I feel more confident with AI, but also less willing to accept the vague hype still flowing around.

Everyone is building AI-powered marketing workflows right now.

AI browsers are all about convenience, but behind the scenes, they're filled with serious security, privacy, and trust issues.

The more abstracted AI becomes, the more valuable it is to understand what’s behind the abstraction.

Why I’m resisting the urge to copy‑paste ML, relearning the math behind algorithms, and building from scratch to bridge marketing and machine learning.

K-Means++ improves K-Means by choosing better initial centroids using a distance-aware probability distribution.

A full walkthrough of building and deploying a SaaS Market Intelligence tool using Streamlit + FastAPI + GPT-4.

ARIMA vs SARIMA explained: the key difference is seasonality.

A detailed, step-by-step explanation of the code from my churn prediction project - covering data cleaning, feature engineering, SMOTE, Random Forest model, and lessons from end-to-end implementation.

Here is how to frame vague business goals into clear machine learning problems using classification, regression, and ranking lenses to make real-world impact.

Despite the hype around deep learning, classical time series models like ARIMA and SARIMA are essential tools for reliable forecasting.

I started data science in 2024, and now 1 year in of masters in DS, building personal projects as well as applying what I learnt into projects for my job I feel there are tons of lessons I can share.

An in-depth walkthrough of my churn analysis capstone project: from data exploration and clustering to actionable insights for customer retention in a competitive e-commerce landscape.

A decade ago, pointers made me quit programming.

A learning log covering project planning for AI/ML, Go language practice, deep learning theory, MLflow basics, and fine-tuning open-source LLMs across Day 5 to Day 8.

Learning Golang structs and pointers while returning to deep learning with PyTorch and prepping for my second year of masters in data science & ML.

Refactored my Go project structure and started the concurrency book — a short but insightful Day 3.

A short, honest reflection on finally understanding pointers in Go — a concept that I picked up after a decade.

Documenting my first day of learning Golang from scratch with real projects and tests.

A reflection on deciding to learn programming properly at 30+ — and why creation, not code, is what brings me back.

Notes on GTM engineering, AI security, data systems, forecasting, backend work, and the technical decisions behind software people actually have to use.