Articles
Long-form, code-first writing on machine learning — from gradient descent from scratch to shipping RAG systems in production.
Shipping Text Embeddings in Production
Chunking strategies, vector store trade-offs, hybrid search and the retrieval evaluation loop nobody talks about.
Vision Transformers vs. CNNs in 2026
When to reach for a ViT, when a modern ConvNet still wins, and how to benchmark the two honestly on your own data.
Prophet vs. Neural Forecasting: A Real Comparison
Structural decomposition versus cross-learning neural models — where each wins, the baseline both must beat, and how to backtest properly.
The Only Loss Function Cheat Sheet You Need
MSE, MAE, Huber, quantile, cross-entropy, focal and contrastive — what each optimises, when it is the right call, and what it does badly.
Feature Engineering Still Beats Deep Learning on Tabular Data
Aggregations, safe target encoding, date decomposition and leakage traps — an evidence-based tour of what actually works on structured datasets.
Why Implement a Machine Learning Algorithm From Scratch
Why bother implementing algorithms yourself when great libraries exist? The case for building your own — what you learn, what it costs, and a five-step method that works.