Cyber Resiliency Labs ← Careers Building AI

Choosing Your Path

Careers Building AI, Episode 6. The capstone. As of mid-2026.

Over this series we walked through the roles. The researchers. The engineers who build on top of models. The people who keep the infrastructure standing, and the ones who break models on purpose to find out where they fail.

You have seen all of it now. One question is left. Which of these is yours?

This article does one thing. It takes everything from the earlier episodes and turns it into a decision you can act on by the end of the week. We are not adding new roles. We are picking one.

Match your background to a role

Most people ask the wrong question. They ask what the best AI job is. The better question is which job fits the person you already are.

Almost everyone reading this has a starting point. You are not breaking in from nothing. You are translating a skill you already have into a new target.

Here is how the common backgrounds map.

If your background is... The roles that fit Why
PhD or research-minded (the hard proofs, the papers, the open questions) Research Scientist, Safety Researcher These usually want a doctorate, or research output that looks like one.
Strong software engineer Research Engineer, Machine Learning Engineer, AI Engineer You have the most doors open of anyone. AI Engineers build products on top of models.
Systems or DevOps (the plumbing) ML Systems engineer, MLOps You already think about reliability and scale, and what happens at 3am when it breaks.
Security AI Red Teamer You break models the way you used to break networks.
QA or measurement mindset (loves a good test) Evaluations (evals) Same instincts, pointed at whether a model actually works.

Find your row. That is your shortlist.

Three ways in

The role is the easy half. The harder half is getting the credential or the proof that gets you hired. There are basically three routes.

Route one: the doctorate

Long and narrow. It opens a specific set of doors, mainly the research roles and a lot of the safety roles. If you want to be the person inventing the next method, this is still the main road.

Route two: a bootcamp or applied courses

The practical fast lane. It can get you into AI Engineer, Machine Learning Engineer, MLOps, and Data Engineering. Pair it with a portfolio and it reaches red-teaming too.

Route three: self-taught, carried by a portfolio

No degree, no bootcamp. Just things you built and can show. This opens AI Engineer, the new Agentic AI Developer role, and red-teaming again. The work speaks louder than the paper.

Notice that two of the three routes need no degree at all. Most of the roles in this series are reachable through applied courses or a portfolio. The doctorate is one door, not the whole building.

Salary and demand at a glance

Money and demand shape the decision too. The figures below are rough and reported, as of mid-2026, and they move fast. Treat them as a map, not a promise.

The pay range is wide. Entry-level applied roles start around $90K a year. At the frontier labs, the very top of research and engineering reportedly reaches into seven figures of total compensation. Where you land depends far more on the role and the employer than on the title.

Role Reported pay band (mid-2026) Reported demand
Agentic AI Developer / agent architect Architects reportedly ~$260K to $420K Sharpest shortage: roughly 8 open roles per qualified person
Safety Researcher Among the highest-paid; frontier bands run high Roughly 4 to 5 open roles per person
Research Engineer (frontier labs) Reportedly into seven figures at the very top Strong, concentrated at the frontier
Machine Learning Engineer Solid mid-to-senior bands Largest gap in sheer numbers: ~1 qualified person per 3.5 openings
MLOps / ML Systems Strong, especially with reliability experience About 2.8 openings per person
AI Engineer (applied) Entry around $90K, rising with product impact Broad and steady across the market

Two patterns stand out. The shortages bite hardest in agentic AI and safety. The highest pay clusters around frontier research engineers, agent architects, and safety.

Spotlight: the Agentic AI Developer

Put pay and demand together and one role keeps jumping out. It barely existed two years ago.

The Agentic AI Developer builds systems that do not just answer. They act. They plan, they call tools, they take steps on their own. It is the fastest-growing and newest role on the map, and demand is running well ahead of supply.

The skills are surprisingly learnable:

The on-ramp is short. A backend engineer can reportedly pivot into this in something like two to four months, because so much of it is software engineering pointed at a new target. Meanwhile hiring is expensive and getting more so. Agent architects are reportedly landing around $260K to $420K, as of mid-2026.

A learning-path framework anyone can follow

Whatever tier you picked, you still need a plan. This one works for every route. Five steps, in order.

  1. Pick a tier. Choose the role family that matched your background. Not the hottest title. The one that fits.
  2. Take one applied course in it. Just enough to get a real feel for the day-to-day before you commit.
  3. Build a portfolio project. One thing that works, that you can show and talk about.
  4. Go deep with an expert program. Do this once you know it is the right fit. This order saves you from sinking months into the wrong path.
  5. Study real job posts. Pull out the exact skills and tools they keep asking for, then aim your project and your study straight at that list. The market tells you what to learn, if you read it.

The full per-role learning paths

Each tier has its own companion article with every step spelled out. Start with the one that matched your row.

Go pick your path

The demand is real across every single tier, from research down to the applied roles. You do not need permission to start.

The fastest way in is to build something and show it. So go build the thing.

Watch Episode 6 on crlab.ca

By Zubair Ashraf · CR Labs · crlab.ca