Articles on Machine Learning, Deep Learning, and Data Science
From linear search to RAG: build intuition behind indexing, ranking, relevance, and semantic search. With Python examples and a didactic mini search engine from scratch.
Turn a Raspberry Pi 5 into a personal AI dev server — Docker, Tailscale, persistent services, and a running demo app. Practical guide starting from your first SSH session.
A practical, complete guide to uv — the modern replacement for pip, venv, pyenv, and pip-tools. From zero to production, with the shortcuts you will use every day.
Practical guide to deploying on AWS with Docker containers. We start with Lambda (the simplest option), move to Fargate for long-running jobs, and build a complete pipeline with enrichment and vector search. Real-world example: a news scraper for a newsletter.
Learn Docker by building: a FastAPI API with an ML model that starts running locally and ends production-ready. Each problem leads to the next concept — Dockerfile, Compose, volumes, networking, security, and deployment.
Learn Terraform by building: a news scraper that starts with manually created resources and ends with versioned, modular, automated infrastructure. Each problem leads to the next concept.
CLAUDE.md, CLAUDE.local.md, rules, skills, agents, hooks, memory, settings — Claude Code has dozens of configuration files. This guide explains each one, where to put it, when to use it, and how they all relate to each other.
Stop stashing, switching, and losing context. Git worktree lets you check out multiple branches simultaneously, each in its own directory, all backed by a single repository. Here is the complete, step-by-step guide.
Context engineering is the discipline that separates hobbyist LLM experiments from production-grade AI systems. This guide covers everything — from RAG and memory to agentic loops, token budgeting, and security — so you can build agents that actually work.
A step-by-step guide to building a fast, bilingual personal blog and portfolio site using Astro 5, Tailwind CSS v4, and Tokyo Night theme — exactly how this one was built.
What if we could eliminate traditional feature engineering by converting raw tabular data into text and then into vector embeddings? Exploring the tab2vec framework and what the experiments revealed.