Prepare for a research or applied-AI role at a frontier lab.
Free lessons and a from-scratch Python book, with the code on Colab.
Free lessons and a from-scratch Python book. Each chapter derives the math, runs on Colab and measures a real model.
Start here: who this is forWhat you get, and the promises we keepOr go straight to Chapter 1. No account neededHow each chapter works
Derive the idea → Run the notebook → Measure a real model
Newsletter
One lesson by email each week.
Free; unsubscribe any time.
Two places to begin
Choose your first project
Project 110 min read
Calibrate an AI judge
Read an AI judge's confusion matrix before you trust its error score.
Try the judge-calibration guide
Requires a few cases you label by hand before the judge sees them.
Project 27 min read
Understand how an agent works
See what an agent remembers after a restart with a small Python example.
Try the memory example
Requires basic Python.
Browse by skill:Measure modelsBuild systems around modelsMake it fast and bigFoundations
Latest
Newest first- EssayA Job ID Is Not a Result: Check the Output Before Marking It DoneA scheduler's job ID, like a deploy hook's 200, says a request was accepted. A five-state model and a small check that marks work done only when the results file exists, from the right checkpoint, with every row.
- NoteReview a Model Upgrade Like a Code Change: Count the Regressions
- GuideA Coding Agent That Writes Its Own Tests Is Grading Itself
- NoteCheck a Codex Fix Against the Original Failing Case
Two ways to learn
Learn with guided lessons
Short lessons with worked examples and checks, in order. The first lesson of every course is open to read.
A free account opens the remaining lessons and saves your progress.
The Prof Rod library
Build Your Always-On AI Agent From Scratch
Rod Rivera
The book, free to read
Understand an agent by building one
Tools, memory, permissions, evaluation and recovery, built in Python from scratch and measured on Claude. No account needed.
Open the bookWhere this skill is used
From the lesson to the work
Job · Anthropic
Research Engineer, Post-Training Model Evaluations
San Francisco, CA or Seattle, WA
Practice the paired comparisons used to investigate changes in post-training evaluation results.
Last checked . View role
Job · Anthropic
Research Engineer, Code RL (Reinforcement Learning)
San Francisco, CA or New York City, NY
Practice the experimental design and interpretation used to evaluate coding and RL experiments.
Last checked . View role
Three characters in the lessons
Each lesson explains a mechanism, builds it, and measures what it can and cannot do.
Prof Rod
Teacher
Explains the mechanism on the board and works the example, so you see the reasoning as well as the result.
Mator
Builder, visiting
Builds exactly what you specified, so gaps in the specification show up fast.
Quackster
Alchemist, visiting
Asks the question you skipped.
Who teaches this
Rod Rivera
Professor of the Practice in AI at ITAM, also teaching at CDTM/TUM. PhD, more than twenty peer-reviewed machine-learning papers, and fifteen years building ML systems at Alibaba Cloud, Huawei and Samsung. The lessons include the experiments that failed.
More about RodGet next week's lesson by email
Newsletter
One lesson a week, free
This only subscribes you to the newsletter; it doesn't create a course account.
