Careers Building AI · Episode 1

The AI Careers Map

"I want to work in AI" is about a dozen different jobs. This guide sorts them out.

Almost everyone I talk to lately says the same thing. They want to work in AI. And I always ask the same thing back: okay, but which job?

The person who invents a new training method and the person who wires a model into a customer support tool both say they work in AI. They're both right. But their days, their skills, and their paychecks have almost nothing in common.

So here's a map. Forget the titles for a second and ask what each job is actually for. There are five answers. I call them tiers.

The five tiers

1. Research

This is where new ideas come from. New architectures, new training methods, things that didn't exist last year. The work gets measured in papers and breakthroughs, and most of it doesn't pan out. That's the job.

2. Engineering

Once a method works, someone has to turn it into a system people can use. Engineering builds and ships the models, and the apps that wrap around them.

3. Infrastructure

These models are famously expensive to train and run. Infrastructure is where the compute bill gets won or lost. It's the plumbing nobody sees and everybody needs.

4. Safety and Red-Teaming

This tier breaks things on purpose and then fixes them. Find the failures before users do, measure whether a model is actually any good, and work on keeping it pointed where we want. It's been growing fast as of mid-2026.

5. Application

Where AI meets a real customer. People here build the features and products on top of existing models, often without training anything from scratch. It's the widest on-ramp into the field.

The 13 roles

Thirteen roles spread across those five tiers. Salary bands below are rough, hedged, and current as of mid-2026. Treat them as a sketch, not a quote.

Entry-level roles across these tiers start around $90K as of mid-2026. The numbers move fast, so don't anchor on any single figure.

Three confusions to clear up

A lot of these titles sound interchangeable. Three pairs trip people up the most.

Research Scientist vs Applied Scientist vs Research Engineer

Think of it as a relay. The Research Scientist asks the open question. The Applied Scientist shapes the answer into something useful. The Research Engineer builds the code that proves it works. Same project, three different handoffs.

MLE vs AI Engineer

The MLE works close to the model itself, training and tuning it. The AI Engineer usually takes a model that already exists and builds the product around it. One shapes the engine. The other builds the car.

Red Teamer vs Safety Researcher

The Red Teamer is the attacker, finding specific holes in a specific system right now. The Safety Researcher works the longer problem: how to build models that behave well in the first place. Same goal, very different time horizon.

The pay spread

The range is wide. As of mid-2026, entry-level roles start around $90,000, and the top research jobs at frontier labs can reach seven figures. Most people land somewhere in the long middle, and where you land depends more on tier and company than on the title on your badge.

Demand is the other half of the story. Across every tier, hiring is running well ahead of supply, and not just for the famous research jobs. The infrastructure people, the safety people, the data engineers: companies want all of them and can't hire fast enough. That's the read as of mid-2026, and it can change.

How to use this guide

Today's piece is the whole map. Each article after this one opens a single tier and goes deep: the real day-to-day, the skills, the training paths in, and a concrete learning path you can follow.

Pick the tier first. The title comes after the fit.

• Researchers — inventing the methods everyone else builds on.
• Builders — turning working models into shipped products.
• Infrastructure — training and serving without melting the budget.
• Defenders — testing, attacking, and aligning the systems.
• Choosing Your Path — picking the room that fits you.

Watch the series: crlab.ca/videos/careers-building-ai/

By Zubair Ashraf · CR Labs · crlab.ca