How a CNC Instructor Built an AI Grading Assistant — and What Happened Next

How a CNC Instructor Built an AI Grading Assistant for Vocational Education

When Adam Hathaway started teaching machine tool technology at Chabot College, grading was one of the most grueling parts of the job. Students in his CNC programming courses submit G-code — the language that tells a machine exactly where to move, how fast to spin, and how deep to cut — accurate to the fourth decimal place. Grading it meant going line by line, manually checking each value against a master program, every semester.

"An enormous part of my life is going through and grading line by line, these codes," Adam said in a recent Nectir webinar on AI in vocational education.

After about a year of using Nectir AI, he's built a practical, evolving system that's changed how his students learn, and how he spends his time.

Can AI be used in vocational and trades education?

Yes, and it looks different than in a traditional academic classroom. For technical instructors like Adam, the challenge isn't just engagement or writing quality, but precision. CNC programming requires students to produce highly specific output: code that must match an instructor's exact standards, values, and methods, with very little margin for error.

Generic AI tools weren't the answer. They pull from the open web, which means students might get responses that don't match how their instructor teaches.

"If I was using ChatGPT, I'm not sure where they're going to grab that information. I'm darn well sure it's not going to be mine."

What Adam needed was an AI grounded entirely in his own course materials. That's what Nectir made possible.

How do you build a custom AI assistant for a CNC course?

Rather than creating one catch-all assistant, Adam built individual Nectir AI assistants for each project in the course. Each one is trained on the relevant master program, his preferred equations, and the specific standards for that assignment.

"Each project has its own different standards. By making a new assistant for each project, that standard is maintained within that project."

The knowledge base for each assistant includes what he calls the "Encyclopedia Machina" — a compendium of machining information, his preferred calculation methods, and the exact values he wants students to use. When a student asks how to calculate RPM, they get Adam's formula, not a generalized answer from the internet.

What does AI-assisted grading look like in a technical course?

The assistant does two main things: it grades student code, and it helps students understand why something is wrong. When a student pastes their G-code into the assistant, it outputs a line-by-line review — written in CNC-appropriate comment formatting, flagging every error with a specific explanation. As Adam detailed, "It gives them very specific information: on this line, this thing is wrong. Work offset number is incorrect. Spindle speed is incorrect. And these are accurate critiques."

Beyond grading, the assistant is available around the clock, which matters for working students.

"The student has access to this AI whenever. If they're working a day shift and midnight is when they get to do their work, they can go ahead and submit these codes and get really good output."

How long does it take to set up an AI assistant for a course?

Adam estimates about 20 hours over the course of a year, most of it front-loaded as he was learning the tool for the first time. His current prompt is 14 pages long, built up iteratively from Nectir's built-in templates. One of his biggest breakthroughs was realizing the AI could help improve itself, "[Nectir AI] not only tells you why it made that mistake; it also fixes itself. I can ask it to edit the prompt so that error doesn't continue."

Each fix gets faster as you learn the system. As Adam put it, the amount of time it takes to make a correction "gets smaller and smaller."

Does using AI in CNC courses actually improve student outcomes?

According to Adam, yes and measurably. "Students who come in knowing nothing about CNC; it's just a firehose of information. Having the AI there helped get over that apprehension about getting marked wrong. The programs I get back look so much better. Even on tests, they understand the coding better. When it comes to reading and writing G-code, it's been a godsend."

From a grading standpoint, the shift has been just as significant.

"The codes I get are so much better. I can blaze through them because there's very little red. That is 100% from using these AIs."

Adam's experience at Chabot College is one example of a broader shift happening across the California Community College system, where Nectir AI is now available to 2.1 million students across all 116 colleges. Vocational and CTE programs, from machining and healthcare, to construction technology, and more, represent some of the highest-stakes, most technically specific learning environments in higher education. They're also some of the most underserved by generic AI tools.

Purpose-built AI infrastructure, grounded in instructor-owned content, is what makes the difference in these contexts.

Adam goes deeper in the full recording — including a live walkthrough of his prompt, a real student running a CNC machine on code reviewed by the AI, and a Q&A on how to get started. Watch the full recording here.

Curious what this could look like in your courses or institution? Book a demo with the Nectir team.