4. AI Engineer
Evidence label: Role family. AI Engineer is the broad application-engineering title; employer scope still ranges from model adaptation to full-stack product delivery.
What they do
Build end-to-end AI systems. Not researchers (they don't train models from scratch), but not just integrators either. They take LLMs and orchestration frameworks and build systems that ship. Think of them as software engineers who've added LLM integration, evals, and AI product intuition to their stack.
Responsibilities
- Design and implement LLM-powered applications (chatbots, agents, pipelines)
- Build evaluation frameworks to measure model output quality
- Integrate AI capabilities into existing software systems
- Monitor AI systems in production (latency, cost, quality drift)
- Select appropriate models for specific tasks (capability vs. cost tradeoffs)
- Implement fine-tuning or RAG when base models aren't sufficient
Required skills
| Technical | Soft |
|---|---|
| Strong software engineering foundations | Product judgment |
| Python (primary), JavaScript (often needed) | Pragmatism over research purity |
| Familiarity with major LLM APIs (Anthropic, OpenAI, Gemini) | Fast iteration mindset |
| Eval design and measurement | Ability to work with ambiguous requirements |
| Understanding of embeddings, RAG, agent frameworks | Communication of AI limitations to stakeholders |
| MLOps basics (deployment, monitoring, versioning) |
The critical distinction from ML Engineer
AI engineers work with existing models. ML engineers build and train models. In practice, most companies hiring in 2025-2026 need AI engineers (apply the models) not ML engineers (build the models).
Entry paths
Software engineer (most common), backend engineer, data engineer, ML engineer transitioning to applied work.