5 AI Roles Paying $90K to $400K: Application Playbooks + 30-Day Roadmaps

You commented ROLES on the carousel. Here's the deep version — what each role actually requires, who's hiring as of May 2026, the exact application playbook per role, and a 30-day learning roadmap that gets you from "I'm interested" to "I'm a credible candidate."

All five roles are verified live in May 2026. None require a PhD. All accept international applicants.

The pay ladder (quick reference)

Role	Pay (US)	Background you need	Key hiring	Apply at
Prompt Engineer at SaaS	$90K-$220K	No formal gate — show measurable improvements	Intercom, Notion, Linear-tier, AI-native startups	Their careers pages
AI Agent Engineer	$120K-$400K	Software dev background	Every SaaS + AI labs	GitHub-first apply
AI Evals Engineer	$130K-$250K	ML or test-engineering	Anthropic, OpenAI, DeepMind, Scale AI	Lab careers pages
AI Red Teamer	$130K-$280K	Cybersecurity or ML	DeepMind, Anthropic, Amazon, Capital One	Security teams
AI Solutions Architect	$145K-$268K	Solutions / senior dev	Accenture, Deloitte, AWS, Azure, GCP	Consultancies first

---

1. Prompt Engineer at SaaS — the lowest-barrier entry point

What you actually do: You productionize prompts inside SaaS apps. The app already calls Claude or GPT — your job is to make those calls cheaper, faster, and more reliable. You build evaluation loops, prompt versioning systems, A/B tests on prompts in production.

Pay reality (verified): $90K entry, $220K senior. Glassdoor 2026 avg $129K. PE Collective: entry $90-125K, mid $130-175K, senior $170-220K.

Application playbook:

1. Pick 3 SaaS companies you actually use. Find their AI-powered feature.
2. Reverse-engineer their prompt by using the product. Note what you'd improve.
3. Write a proposal: "Your X feature does Y. Here's a prompt that does Y better, with this evaluation methodology."
4. Send it to their head of product or AI lead. Cold outreach via LinkedIn DM beats application portals.

30-day roadmap:

• Week 1: Read OpenAI's prompt engineering guide + Anthropic's prompt engineering docs end-to-end.
• Week 2: Build 5 evals for a real task. Use OpenAI Evals or Anthropic Inspect.
• Week 3: Pick your 3 target companies. Study products. Write proposals.
• Week 4: Reach out. 5-10 senior-product-leader DMs per week.

Common mistake: Applying through the careers portal. This role hires through portfolio + cold outreach.

---

2. AI Agent Engineer — the highest ceiling ($400K)

What you actually do: Build multi-step autonomous AI workflows. LLM calls tool, tool returns data, LLM decides next step. Frameworks: LangGraph, CrewAI, Claude Agent SDK, AutoGen.

Pay reality (verified): $120K-$400K total. ZipRecruiter US avg $128K; SecondTalent reports $80-250/hr globally. Mid-level $120-180K, senior $200K+, frontier-lab senior pushing $400K total comp.

Application playbook:

1. Pick a real task someone does manually that an agent could automate. Not a toy problem.
2. Ship the agent publicly on GitHub. README shows before/after.
3. Write up the architecture decisions: framework choice, failure modes, fixes.
4. Apply with the GitHub repos as your CV. The work IS the credential.

30-day roadmap:

• Week 1: Build a basic LangGraph or Claude Agent SDK agent — one multi-step task.
• Week 2: Build a second agent using 2+ tools (web search, code execution, file I/O). Ship on GitHub.
• Week 3: Build a third agent solving a real workflow. Document failure modes.
• Week 4: Apply. Lead with the 3 GitHub repos. Write a thread/post on what you learned.

Common mistake: Building toy agents and never shipping. Companies want code in the wild.

---

3. AI Evals Engineer — the discipline of the 2026 AI stack

What you actually do: Design test suites that prove a model works. Build benchmarks, eval datasets, regression tests. Frontier labs spend more on evals than training data now.

Pay reality (verified): $130K-$250K. JobCannon reports entry $130-173K base, with equity $143-208K. SF/NYC: $150-208K.

Application playbook:

1. Pick a public eval framework: HELM, Inspect (UK AISI), OpenAI Evals.
2. Reproduce one published eval from scratch. Run it on Claude / GPT / Llama.
3. Add a new eval — a domain or failure mode it doesn't cover.
4. Apply to eval teams at Anthropic, OpenAI, DeepMind, Scale AI with reproduction + new eval as portfolio.

30-day roadmap:

• Week 1: Pick HELM or Inspect. Read the methodology paper.
• Week 2: Reproduce one eval. Run on 2-3 frontier models. Compare to published results.
• Week 3: Design your own eval for a capability gap (long-context retrieval, agentic reasoning, refusal calibration).
• Week 4: Apply.

Common mistake: Treating evals as "writing test cases." Real evals are research. Methodology over coverage.

---

4. AI Red Teamer — the safety-engineering boom

What you actually do: Break AI models before they ship. Adversarial attacks, prompt injection, jailbreaks, policy bypass. You stress-test, the company patches.

Pay reality (verified): $130K-$280K. DeepMind Senior Security Engineer (Agentic Red Team) $166K-$244K. Amazon Senior Security Engineer (AI Red Team) $178K-$226K. Lead roles $200K-$300K+.

Application playbook:

1. Pick HackTheBox AI Red Team path or OWASP LLM Top 10 as your curriculum.
2. Document 3-5 red-team findings on public LLMs. Clean, reproducible exploit reports.
3. Publish on a blog. Tag model providers on X when relevant.
4. Apply to AI security teams at Anthropic, OpenAI, DeepMind, Amazon, Capital One, Microsoft.

30-day roadmap:

• Week 1: OWASP LLM Top 10 reading. Set up API testing environment.
• Week 2: Run 5 jailbreak / prompt-injection attempts. Document what worked.
• Week 3: Pick one finding worth a writeup. Reproducible exploit + defense proposal.
• Week 4: Publish + apply.

Common mistake: Treating it like CTF. AI red teaming is about model behavior + alignment, not classical web exploitation.

---

5. AI Solutions Architect — the enterprise gateway

What you actually do: Design how a company integrates AI end-to-end. RAG architecture, vector DB choice, model selection, fine-tuning vs. prompt, security + compliance design.

Pay reality (verified): $145K-$268K. Glassdoor avg $210K. ERI SalaryExpert $180K. Entry $126K, senior $200K, principal $250K+.

Application playbook:

1. Get AWS AI Practitioner or Azure AI Engineer cert (60-80 hours of study).
2. Build 1-2 integration portfolio projects: a RAG system on real documents, an agent for a business tool.
3. Document architecture decisions (model choice, vector DB, eval strategy).
4. Apply to consultancies first (Accenture, Deloitte AI practice). Cloud-provider solutions teams next (AWS, Azure, GCP).

30-day roadmap:

• Week 1: Start the AWS AI Practitioner or Azure AI Engineer cert prep.
• Week 2: Build a RAG system on a real dataset. Document the architecture.
• Week 3: Build a tool-using agent for a business workflow.
• Week 4: Take the cert exam. Apply with cert + 2 portfolio projects.

Common mistake: Cert without building. Cert + 0 projects = CV. Cert + 2 working systems = hire.

---

Which role should YOU pick first?

• No coding background, want to start tonight: Prompt Engineer at SaaS.
• Software dev with 2+ years, want max ceiling: AI Agent Engineer.
• ML/data background, methodology-minded: AI Evals Engineer.
• Cybersecurity / pentesting background: AI Red Teamer. Fastest pivot of the five.
• Solutions architect, consultant, senior dev: AI Solutions Architect.

You can apply to multiple — they're not exclusive. Many people stack 2-3 (e.g., Prompt Engineer first for income, then Agent Engineer when GitHub is strong enough).

What to do this week

1. Today: Pick one role. The one that matches your strongest 2 existing skills.
2. Within 7 days: Start the 30-day roadmap. Schedule the time.
3. Within 30 days: Ship the portfolio asset. Submit your first applications.

The bar to APPLY is lower than the bar to GET IN. Most viewers will save this carousel and never start the roadmap. Be the 1 in 100 who does Week 1.

Follow @africantechbro for more. Tag me when you submit your first application — I want to see the screenshot.