
How to Do Evals in 2026
A tool-agnostic framework for evaluating AI agents
Practical writing on building, testing, and operating AI systems, workflows, and automation, based on my projects and experiments. Subscribe to get ideas you can reuse in your own work.

A tool-agnostic framework for evaluating AI agents

Part 3 of AI Dev Tools Zoomcamp

The 5-step framework that will land you a job

Part 2 of the AI Dev Tools Zoomcamp series

Serving 100k+ Learners with a Serverless FAQ Assistant

An Analysis of 113 AI FDE Job Postings
Part 1 of the AI Dev Tools Zoomcamp series

16 sections your project should have in the README

How I release Python packages with agents and without

When preparing personal plans for AI Shipping Labs members, I often get the question, “What should I build for my portfolio?” There are too many options, and no clear way to pick one.

But I learned a ton while building them

I sometimes run offline workshops.

My workflow for AI-generated interfaces: how I use mockups and design guidelines to avoid AI slop

Start with one real task, then add skills, subagents, and extra tools when you need them

How a small in-process Python library replaced Elasticsearch in my RAG notebooks that I use for teaching.

A remote server, tmux, voice-driven Claude Code, and a workflow designed around the in-between time

I consolidated all the GenAI workshops I've ever taught into one place. Here's what's inside and how to take them.

How a niche bedtime-book project became a system for generating chapters, audio, EPUB files, and Amazon KDP print artifacts.
This is what the students of the second cohort built + AI Shipping Labs new

At DataTalks.Club, we run free courses like ML Zoomcamp, Data Engineering Zoomcamp, MLOps Zoomcamp, and LLM Zoomcamp.