Built In puts AI engineer total compensation at $211,243, with average base pay of $184,757 and a median base of $180,000. Total compensation is the honest way to read that number, because it includes more than what lands in your bank account each month, and plenty of posts online quote it as if it were salary. Source: Built In, AI engineer salary.
Numbers 1 and 3 cost nothing. Number 2 has no published price at all. Numbers 4 and 5 are the two to budget for, and one of them changes price this month.
Cost: free. Time: Harvard's own words are "the course's ten weeks of material". You get: a free CS50 certificate if you score 70% or better on every problem set and complete the final project.
Everything else on this list assumes you can write Python, so this is where you start if you cannot yet. It is the best free programming course there is, the problem sets are properly designed, and the certificate is issued by CS50 itself at no cost, which is rare. You do need a free edX account to submit your work, and the separate edX verified certificate is a paid product you can ignore.
Start here: CS50's Introduction to Programming with Python. The certificate rules are on the certificate page, and they are worth reading before week one so you know what you are aiming at.
Cost: no published price. Exam: AI-103, 120 minutes, proctored through Pearson VUE. Expected: Python experience. You get: a Microsoft credential whose role tag on the page is literally "AI Engineer".
This is the first credential on the list that a recruiter's filter recognises by name, and Microsoft's own page tags the role as AI Engineer, which is exactly the phrase you want sitting next to your name. It is a real proctored exam rather than a course completion, and it tests building AI apps and agents on Azure, so it pairs well with a project you have actually shipped.
Details and booking: Azure AI Apps and Agents Developer Associate.
Cost: the course is free, the API usage is not. Time: 9 hours, 67 lessons. You get: a completion badge. Needs: Python proficiency, basic JSON, and an Anthropic API key.
This is the one that turns "I have done a Python course" into "I have built something that talks to a model". Sixty-seven lessons is a lot of hands-on, and it is the shortest route from knowing the language to holding a working project you can put in front of someone. Do it after CS50, because it does expect you to already be comfortable in Python and to read a JSON payload without flinching.
Start here: Building with the Claude API.
Cost: Coursera states "A Coursera subscription costs $49 / month", with financial aid available. Time: "2 months at 10 hours a week", across 3 courses. You get: a shareable specialization certificate.
This is the one with the university name on it, and it is where the words people throw around in interviews stop being vague. Three courses, two months at ten hours a week, and by the end you can say what a model is actually doing rather than describing it as magic. Because it is a subscription rather than a one-off fee, the cost genuinely depends on your pace, which is an argument for booking the ten hours into your week properly.
Course page: Machine Learning Specialization. If the fee is the blocker, apply for financial aid before you enrol rather than after.
Cost: $75 USD at beta pricing. Exam: 85 questions, 170 minutes, proctored. You get: the AWS associate-level credential for machine learning engineering.
This is the heaviest credential here and it goes last, because it assumes everything above it. It is also the one with a clock on it, so read this part carefully before you book anything.
Exam page: AWS Certified Machine Learning Engineer Associate.
The reason people buy a certification and never sit it is that the page gives you hours, not a schedule. Paste this into Claude once per certificate and it hands you back a week-by-week plan built around the time you actually have.
I am working towards [certificate or course name] and here is the official page describing it: [paste the URL, plus anything you copied off the page about length, format, modules and exam topics].
I have [number] hours a week and I want to be done by [date]. Before you plan anything, ask me up to three questions about how much Python I actually write, what I have built before, and whether I have used a cloud platform, so you do not waste weeks teaching me things I already know.
Then build me a week-by-week plan. For each week give me the specific topics, roughly how long each session should take, and one small thing I have to produce by the end of that week that proves I understood it, like a working script or a short explanation in my own words. Flag the two or three topics most likely to catch me out and schedule those early rather than last. Finish with a checklist for the final week before I sit it.This is the part that does the hiring. A certificate says you attended, a working project says you can build. Run this once after each item on the list and you finish the five holding a portfolio rather than a list of names.
I have just finished [course or certificate name] and here is what it covered: [paste the module names or syllabus from the page]. Here is my situation: [what you can write in Python, what you have built before, how many hours you have, and one industry or hobby you actually care about].
Pick ONE small project that uses what this course taught me, is genuinely finishable in the hours I have, and would make sense to someone hiring for an AI engineering role. Do not give me a list of options, choose the best one and tell me why you chose it over the obvious alternative.
Then give me the file structure, the exact steps in order, and the first working version of the main file with comments explaining each part, so I can run it today and see something happen. Call out where I will need an API key or a cloud account, and what it will cost me. Finish with the two sentences I would use to describe this project on a CV.If you have been looking into this, you have almost certainly seen Hugging Face's AI Agents Course recommended, and it deserves it. The reason it is not here is that it already anchors two of my other guides, and I would rather hand you five things you have not seen from me before than repeat myself. The facts on it are fine. It just cannot carry a third list.