Modelcraft
Engineer's tool bench representing craft and deliberate practice
PLATE 01

What you get that most AI courses leave out.

Modelcraft teaches method, not narrative. This page describes what that means in practical terms.

Back to Home
PLATE 02

Six things that matter when choosing a course

Written, specific feedback

Every reviewed artefact gets written comments tied to what you actually produced. Not a score. Not "good work." Something you can act on.

Work-order structure

Each unit states inputs, method, output and hours before you start. You know exactly what you are taking on.

Honest prerequisites

Each course lists what you need to know, plainly. The six-month programme includes a technical conversation before enrolment confirms.

One-to-one office hours

Office hours in the practitioner programme are individual slots. If you are stuck on something specific, the answer is specific — not a group FAQ.

Left out on purpose

Every course page lists topics it deliberately does not cover. This prevents enrolling on a false premise.

Taught by working engineers

Teaching staff work in the field currently. The content stays close to practice because that is where the teachers spend their time.

PLATE 03

What each benefit means in practice

Expertise in production AI work

The instructors have built things in production and carry that into their teaching. The failure taxonomy material in the prompt engineering course comes from a real incident log. The retrieval systems content in the practitioner programme was written by someone who runs them. This is not a universal feature of AI education.

Process you can reuse

The reading protocol from the paper-reading course is yours to keep. The prompt-versioning reference repository and regression harness template from the engineering course go into your own repositories. Artefacts from the practitioner programme are in repositories you own. The deliverables are designed to be reused, not to demonstrate completion.

Support that answers the actual question

Cohort channels provide day-to-day contact. Individual office hours provide focused time on specific problems. Written review provides the most transferable form of feedback. These are three different things, and all three are included where they are relevant to the course.

Transparent fees

All fees are listed on the course pages before you contact us: RM 640, RM 1,780, RM 4,700. The practitioner programme includes compute credit. There are no hidden materials costs, no upsells, and no payment is collected before enrolment is confirmed.

Outcomes that are honest to describe

Modelcraft makes no claim about employment or earnings. What we can describe honestly: you will have worked through a body of material with a practising engineer watching and commenting, and you will have produced artefacts you can show to someone who asks what you can do. That is what the courses produce.

PLATE 04

How this compares with other approaches

FEATURE TYPICAL AI COURSE MODELCRAFT
Feedback on your work Automated quiz scores Written line-by-line review
Prerequisites stated Vague or absent Listed plainly on every course page
Office hours Group forum or none One-to-one slots (practitioner programme)
Scope honesty Everything sounds covered "Left out on purpose" section on each page
Teaching staff May not work in the field Practising engineers, currently working
Deliverables you keep Completion certificate Reusable protocols, repos, artefacts
Fees stated upfront Often hidden until checkout Listed on every course page in RM
PLATE 05

What only Modelcraft does

Work-order format for every unit

The work-order structure — input, method, output, hours — runs through every unit of every course. It is not a framework used by other schools and it comes from the instructors' own engineering practice.

Pre-enrolment technical conversation

The six-month programme involves a short technical conversation before a place is offered. This is so both sides can decline without awkwardness — and because a poor fit wastes twelve weeks of your working life.

Incident log as course material

The failure taxonomy in the prompt engineering course is drawn from a real incident log kept by one of the instructors. This is not case-study pedagogy; it is the actual record.

Team cohort option with your own product

The prompt engineering course can run as a team cohort where participants work on their own product rather than a toy example. Reviews are written against your actual codebase, not a generic starter project.

PLATE 06

Where we are now

3
courses running since 2022
140+
developers enrolled across all cohorts
12
max seats per prompt engineering cohort
24
weeks in the practitioner programme
100%
of sessions recorded and available to participants
PLATE 07

Not sure which course fits your situation?

Say so in the message. Describing your background and what you are trying to learn is enough to start a useful conversation.

Send an Enquiry