Data Science
A working engineer's guide: no ceremony, current tooling, and every topic ends with something running on your machine. Written for practitioners coming from infrastructure and security who want data skills that survive contact with production.
The Field, Mapped
Analytics, machine learning, and AI engineering are three different jobs. What each one actually does in 2026, and where to aim.
FoundationsThe Modern Python Stack
uv for environments, notebooks vs. scripts, and a project layout that won't embarrass you in six months.
FoundationsData Wrangling: Polars, Pandas, and Parquet
Eighty percent of the job is shaping data. The split-apply-combine mindset, columnar formats, and when DuckDB replaces a cluster.
Core SkillsStatistics That Earn Their Keep
Distributions, uncertainty, hypothesis tests, and the Bayesian intuition — the minimum honest toolkit for claims about data.
Core SkillsThe Solitaire Solution: Stanisław Ulam and the Genesis of Monte Carlo Simulation
How a recovering mathematician's hospital card game became the Monte Carlo Method — and why the gamblers of 1913 lost everything to the same probability they refused to understand.
Probability & SimulationVisualization and Storytelling
Charts are arguments. Making ones that persuade without lying, and dashboards that people actually read.
Core SkillsClassical Machine Learning Still Wins on Tables
scikit-learn, gradient-boosted trees, cross-validation done honestly, and why boosted trees outlive every hype cycle on tabular data.
Machine LearningDeep Learning with PyTorch
Tensors, autograd, and training loops from first principles — plus when deep learning is the wrong tool and a boosted tree would do.
Machine LearningWorking with Large Language Models
Prompting as engineering, retrieval-augmented generation, fine-tuning vs. context, local models, and evaluation that isn't vibes.
AI EngineeringData Science for Security Operations
Anomaly detection on network telemetry, log analytics at scale, and why precision beats recall when every alert costs an analyst's hour.
AppliedShipping It: MLOps Without the Buzzwords
Versioning data and models, monitoring drift, and the boring reliability engineering that separates a demo from a product.
Production