The AI Jobs Landscape

The job title that didn't exist three years ago

In early 2023, a company called Anthropic posted a job listing for a "Prompt Engineer and Librarian." The salary range: $175K-$335K. The internet lost its mind. How could writing instructions to an AI pay more than most engineering roles?

But the job wasn't about writing sentences. It was about understanding how language models process instructions, testing systematically across edge cases, building prompt libraries that hundreds of internal users would rely on, and measuring output quality at scale. It required a blend of technical understanding, communication skill, and relentless iteration that's genuinely rare.

That role was a signal of something bigger: AI wasn't just creating new tools — it was creating entirely new career paths with compensation that reflects just how scarce the right skills are.

The six AI career families

AI jobs aren't a single category — they span a spectrum from deeply technical to entirely strategic. Understanding the landscape helps you find where you fit.

Role 1: Prompt Engineer

What they do: Design, test, and optimize prompts that get reliable results from AI models. They build prompt libraries, create evaluation frameworks, and work with product teams to integrate AI into user-facing features.

What it really looks like day-to-day:

• Testing hundreds of prompt variations to find what works consistently
• Building evaluation rubrics to measure AI output quality
• Documenting prompt patterns and best practices for teams
• Working with developers to embed prompts into applications
• Staying current with model updates that change how prompts perform

Role 2: AI Trainer / RLHF Specialist

What they do: Provide human feedback to improve AI models. They evaluate AI outputs, rank responses by quality, identify errors and biases, and write examples of ideal outputs. This is the "human" in "reinforcement learning from human feedback."

What it really looks like day-to-day:

• Rating AI responses on accuracy, helpfulness, and safety
• Writing detailed explanations of why one output is better than another
• Identifying subtle factual errors, biases, or harmful content
• Creating training data in specific domains (medical, legal, financial)
• Working with ML engineers to translate feedback into model improvements

Role 3: MLOps Engineer

What they do: Build and maintain the infrastructure that makes AI models work in production. Think of them as the DevOps engineers of the AI world — they handle deployment, monitoring, scaling, and reliability.

What it really looks like day-to-day:

• Setting up model training pipelines and deployment infrastructure
• Monitoring model performance in production (latency, accuracy, drift)
• Managing GPU clusters and compute costs
• Building CI/CD pipelines for model updates
• Debugging production failures when models behave unexpectedly

There Are No Dumb Questions

What's the difference between an ML engineer and an MLOps engineer?

An ML engineer builds the models — they choose architectures, train models, and optimize performance. An MLOps engineer deploys and maintains the infrastructure those models run on — they handle scaling, monitoring, and reliability. Think of it like the difference between a software developer (writes the app) and a DevOps engineer (keeps it running in production). Some roles combine both.

Do I need a PhD for any of these roles?

For ML engineering at top AI research labs (OpenAI, DeepMind, Anthropic research teams), a PhD helps significantly. For MLOps, prompt engineering, AI training, AI PM, and AI ethics roles, a PhD is not required. Practical skills, projects, and domain expertise matter more.

Role 4: AI Product Manager

What they do: Decide what AI features to build, why, and for whom. They sit at the intersection of technology, business, and user experience — translating AI capabilities into products people actually want.

What it really looks like day-to-day:

• Evaluating which problems are good fits for AI solutions (and which aren't)
• Writing product requirements that account for AI's probabilistic nature
• Defining success metrics that work for AI (accuracy, latency, user satisfaction)
• Managing stakeholder expectations about what AI can realistically do
• Working with ML engineers, designers, and business teams simultaneously

Role 5: AI Ethics / Trust & Safety

What they do: Ensure AI systems are fair, safe, and used responsibly. They develop policies for AI use, audit models for bias, handle AI incidents, and work with regulators on compliance.

What it really looks like day-to-day:

• Auditing AI systems for bias across gender, race, age, and other dimensions
• Writing AI usage policies for the organization
• Reviewing new AI features for potential harm before launch
• Monitoring AI incidents and leading response when things go wrong
• Tracking regulatory developments (EU AI Act, NIST AI Framework, state-level legislation)

Role 6: AI Solutions Consultant

What they do: Help organizations identify, evaluate, and implement AI solutions. They bridge the gap between what AI can technically do and what a specific business actually needs.

What it really looks like day-to-day:

• Assessing client workflows to identify AI opportunities
• Building business cases with ROI projections for AI investments
• Evaluating AI vendors and tools for specific use cases
• Managing AI implementation projects from pilot to production
• Training client teams to use AI tools effectively

There Are No Dumb Questions

Which of these roles pays the most?

ML engineers and AI PMs at major tech companies command the highest compensation — $250K-$500K+ at FAANG and top AI labs (including equity). But high salary correlates with high technical requirements and fierce competition. Prompt engineering and AI consulting offer strong compensation ($150K-$200K+) with lower technical barriers.

Are these roles only at AI companies?

No — and that's the opportunity. Every large company (banks, hospitals, retailers, manufacturers) is hiring for AI roles. These "non-AI companies" often have less competition for candidates and offer the chance to have higher impact. A prompt engineer at JPMorgan might be one of 5 people doing that work; at OpenAI, they're one of 50.

Compensation landscape

Salary data aggregated from LinkedIn Salary, Glassdoor, and levels.fyi. Ranges vary significantly by company size, location, and seniority. Verify current figures before making career decisions.

How to actually get hired

The AI job market is competitive. Here's what separates candidates who get hired from those who don't:

Key takeaways

• AI careers span six main families: prompt engineering, AI training, MLOps, AI PM, AI ethics, and AI consulting
• Median AI role salaries range from $85K (AI trainers) to $200K+ (ML engineers and senior AI PMs)
• You don't need a PhD for most AI roles — domain expertise + AI skills is the winning combination
• Non-AI companies (banks, hospitals, retailers) offer less competition and high impact for AI professionals
• Building a portfolio of AI projects is the single most effective way to get hired
• Specialize your story — "I built X for Y industry and achieved Z result" beats generic AI interest