Careers · 8 roles
Eight new AI-native roles.
When AI agents do part of the work, new jobs appear around them. Here is what each one owns, what it does, and where to learn it on this site.
These job titles are new, and companies use them in different ways. Treat them as roles you will see, not as official job categories. We make no promises about jobs or pay.
See sample certificates → Finish a roadmap and you earn one you can download and share.
Agentic engineer
Full learning pathBuilds software by directing AI agents, and keeps the quality high.
- Owns
- The build: features, tests and releases.
- Does
- Tells agents what to write and what to review. Checks their work with tests and human review.
- Why it matters
- A small team can ship much more. This only works if there are strong quality checks.
- Break a feature into small tasks an agent can do
- Review the code the agents write
- Write the tests that catch bad changes
- Fix the things agents get wrong
Our Agentic AI Engineer and Architect paths cover this from first principles to design and launch.
13 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from 2 courses:
- Agentic AI Engineer9 of 9 lessons
- AI Engineering & Agentic AI Architect4 of 15 lessons
Agentic orchestrator
Full learning pathDesigns and runs teams of agents so the work arrives on time and well checked.
- Owns
- The delivery pipeline: how work moves through agents and checks, and how it is released.
- Does
- Arranges agents into workflows. Sets limits, approvals and tests. Watches the system once it is live.
- Why it matters
- One agent is a demo. Many agents, run well, can deliver dependable work at scale.
- Choose how agents hand work to each other
- Set budgets, step limits and approval points
- Read traces to find where a run went wrong
- Decide when a change is safe to release
Our Multi-Agent Orchestrator and Agent Ops paths are built for this role.
12 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from 2 courses:
- Multi-Agent Orchestrator8 of 9 lessons
- Agent Ops & Governance4 of 10 lessons
Forward deployed engineer
Full learning pathWorks with a customer to turn a messy problem into an AI solution that runs in their world.
- Owns
- The customer's problem, the success numbers and the result.
- Does
- Turns vague needs into a clear spec. Builds and adapts a solution. Connects it to the customer's systems.
- Why it matters
- It keeps the project tied to a business result, instead of a clever demo nobody uses.
- Talk to the people who have the problem
- Write down what success looks like in numbers
- Build a prototype on their real data
- Connect it to their tools and hand it over
Our Forward Deployed Engineer path covers a whole customer project, from the first talk to handover. Real customers will teach you the rest.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- Forward Deployed Engineer10 of 10 lessons
AI transformation strategist
Full learning pathDecides where an organisation should use AI, and turns that plan into funded projects.
- Owns
- The AI vision, the roadmap and the business case.
- Does
- Picks the use cases. Turns strategy into projects with a budget. Manages risk and change.
- Why it matters
- It ties AI spending to results that can be measured.
- Find the work where AI would pay off
- Write the business case with real numbers
- Set the rules for risk, data and approval
- Run small pilots and decide what to scale
Our AI Transformation Strategist path covers use cases, business cases, risk, change and pilots. Your own industry knowledge makes it stronger.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- AI Transformation Strategist10 of 10 lessons
AI evaluation engineer
Full learning pathProves whether an AI system is good enough, and keeps checking after launch.
- Owns
- The test sets, the quality scores and the release gate.
- Does
- Builds test cases from real work. Scores answers by rule and by human review. Blocks changes that make quality worse.
- Why it matters
- Without tests, nobody knows if a change made the AI better or worse.
- Collect real cases and write the good answers
- Score results and study the failures
- Add a new test whenever a bug is found
- Report quality, cost and speed before each release
Our AI Evaluation Engineer path covers the whole job, from rubrics and test sets to release gates, monitoring and interview practice.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- AI Evaluation Engineer10 of 10 lessons
AI security engineer
Full learning pathProtects AI systems from attacks, data leaks and misuse.
- Owns
- The threat model, the guardrails and the attack tests.
- Does
- Finds ways to trick or abuse the system. Limits what tools and data an agent can reach. Tests attacks before release and watches for them after.
- Why it matters
- AI reads text it cannot trust and can take actions. That opens new ways to attack a system.
- Map where untrusted text enters the system
- Try prompt injection and data theft on a test copy
- Limit tool access, keys and permissions
- Review incidents and fix the cause
Our AI Security Engineer path covers threat models, the main attacks, agent security, red teaming, monitoring and incident response. Network and cloud security basics are not taught here.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- AI Security Engineer10 of 10 lessons
AI governance and risk lead
Full learning pathSets the rules for safe AI use and checks that people follow them.
- Owns
- The AI policy, the risk register and the approval process.
- Does
- Sorts projects by risk. Reviews vendors. Plans for incidents. Keeps records for audits.
- Why it matters
- Clear rules let teams move fast without hidden risk, fines or bad press.
- Review new AI projects and place them in a risk tier
- Ask vendors about data use, storage and exit
- Keep the register of live AI systems up to date
- Run a practice session for an incident
Our AI Governance & Risk Lead path covers inventory, policy, controls, laws, privacy, fairness, vendors, assurance and incidents. It is not legal advice, and your regional rules are not covered in detail.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- AI Governance & Risk Lead10 of 10 lessons
AI product manager
Full learning pathDecides what AI product to build, for whom, and how to measure it.
- Owns
- The product plan, the success numbers and the launch.
- Does
- Finds user problems. Chooses where AI helps. Writes clear requirements and test cases. Works with engineers and reviewers.
- Why it matters
- AI products behave differently from normal software. Answers vary and need testing, so the plan must include quality and cost.
- Talk to users and watch how they work
- Score ideas for value, feasibility, data and risk
- Write the spec and the test cases
- Review pilot results and decide what to do next
Our AI Product Manager path covers the job from problem framing to launch, metrics and interviews. Deep interface design and user research are not taught in full.
10 lessons in the roadmap. You do not need to finish whole courses, only these lessons, picked from one course:
- AI Product Manager10 of 10 lessons