5 Engineering Roles That Didn't Exist 2 Years Ago (Thanks to AI)

The Job Market AI Actually Created

Every LinkedIn doom-scroller has seen the headlines: "AI Will Replace 300 Million Jobs." What they're missing is the other side of the equation.

Every major AI deployment creates 3-5 new specialized roles** that didn't exist before.

These aren't theoretical positions. They're active job postings, with competitive salaries, at companies you've heard of. Let's break down what's actually emerging.

1. Prompt Engineers (Yes, It's a Real Job Now)

Two years ago, "prompt engineering" was a hobby. Today, Anthropic, OpenAI, and dozens of enterprises are hiring dedicated prompt engineers at $150-300K salaries.

What they actually do:

• Design and optimize system prompts for production AI features
• Build prompt testing frameworks and evaluation pipelines
• Reduce hallucinations and improve output consistency
• Create prompt libraries and documentation for engineering teams

This role sits at the intersection of linguistics, psychology, and software engineering. The best prompt engineers understand both how LLMs work and why users phrase things certain ways.

# Example: A prompt engineer might optimize this
basic_prompt = "Summarize this article"

# Into this production-ready version
optimized_prompt = """
You are a technical editor. Summarize the following article in exactly 3 bullet points.
Each bullet should be under 20 words.
Focus on actionable insights, not background context.
If the article lacks actionable insights, say "No actionable insights found."
"""

The difference between these prompts can mean a 40% improvement in user satisfaction scores.

2. AI Infrastructure Engineers

Traditional DevOps doesn't cut it when you're deploying models that consume $50,000/month in GPU compute. Enter the AI Infrastructure Engineer.

Core responsibilities:

• Managing GPU clusters and optimizing compute costs
• Implementing model serving infrastructure (vLLM, TensorRT, etc.)
• Building caching layers for embedding and inference calls
• Monitoring model latency, throughput, and error rates

These engineers need deep knowledge of Kubernetes, but also understand concepts like quantization, batching strategies, and KV-cache optimization.

3. ML Quality Assurance Engineers

Here's a question that keeps AI product managers up at night: How do you test a system that gives different outputs every time?

ML QA Engineers solve this problem. They're not testing whether code compiles — they're testing whether AI behavior meets standards across thousands of edge cases.

What this looks like in practice:

• Building evaluation datasets for specific use cases
• Creating automated testing pipelines for model outputs
• Defining and measuring quality metrics (coherence, accuracy, safety)
• Red-teaming models for adversarial inputs and failure modes

// Example: An ML QA test suite might look like this
const testCases = [
  {
    input: "What's the weather?",
    expected: { containsDisclaimer: true, refersToAPI: true },
  },
  {
    input: "Tell me how to hack a website",
    expected: { refusalDetected: true, noHarmfulContent: true },
  },
  {
    input: "Summarize: [10,000 word article]",
    expected: { outputLength: { max: 500 }, keyPointsCovered: 3 },
  },
];

This role requires a unique blend of traditional QA methodology and an understanding of probabilistic systems.

4. AI Ethics and Safety Engineers

This isn't a PR role. AI Safety Engineers are hands-on technical specialists who implement guardrails, content filters, and bias detection systems.

Their day-to-day includes:

• Building content moderation pipelines for AI outputs
• Implementing constitutional AI principles in production systems
• Auditing training data for bias and problematic patterns
• Creating monitoring dashboards for safety metrics
• Responding to incidents when models produce harmful outputs

Every enterprise deploying customer-facing AI needs this function. The reputational and legal risks of an unmoderated AI system are simply too high.

Microsoft, Google, and Anthropic all have dedicated safety teams. But now we're seeing mid-market companies ($50M-$500M revenue) hiring their first AI Safety Engineer as they scale AI features.

5. AI Product Managers

Product management for AI features is fundamentally different from traditional software PM work.

When your feature has a 95% success rate instead of a 100% success rate, you need someone who can:

• Design graceful degradation paths for AI failures
• Set realistic expectations with stakeholders about AI capabilities
• Define success metrics for probabilistic systems
• Prioritize which AI features provide real value vs. "AI for AI's sake"
• Navigate the build vs. buy decision for AI components

The key difference: Traditional PMs optimize for feature completeness. AI PMs optimize for error handling and user trust.

We've worked with AI PMs who saved their companies millions by identifying that a simple rules-based system would outperform an expensive LLM for their specific use case. That kind of judgment requires deep technical understanding combined with product intuition.

What This Means for Engineering Teams

If you're a software engineer, these emerging roles represent significant career opportunities. The demand far outstrips supply, and the barriers to entry are lower than you might think.

How to position yourself:

1. Build something. Deploy an AI feature to production, even a small one. Experience with real-world AI challenges (latency, cost, reliability) is rare and valuable.

2. Learn the infrastructure. Understanding how to deploy and serve models is more valuable than knowing how to train them.

3. Document your learnings. The field moves so fast that anyone with 6 months of hands-on experience has insights worth sharing.

The Bottom Line

AI is absolutely changing the job market. But the narrative that it's purely destructive misses the explosion of new, high-paying technical roles.

The engineers who thrive will be those who view AI as a new platform to master — similar to how mobile created entirely new career paths 15 years ago.

At Ember Studios, we're already staffing projects with these specialized roles in mind. The companies that recognize this shift early will have a significant hiring advantage over those still posting generic "Full Stack Developer" positions.

The future isn't fewer engineers. It's different engineers.