The 2025 AI job landscape
The phrase “AI job” now spans wildly different work, salaries, and skill stacks. A research scientist at a frontier lab and a startup engineer wiring GPT into a support tool both have “AI” in the title, but almost nothing in their day matches. This guide separates the real roles, the rough money, and the skills that actually get you hired — so you can target a specific lane instead of “getting into AI” in the abstract.
The core roles and what they pay
AI engineer — builds products on top of foundation models: RAG pipelines, agents, tool use, prompt design, and evaluation. Heavy on software engineering, light on training. Typical range: $130k–$220k in the US, £60k–£110k in the UK.
ML engineer — trains, optimises, and deploys models. Owns data pipelines, feature stores, training infrastructure, and serving. Needs solid maths plus strong engineering. Range: $140k–$240k US, £65k–£120k UK.
Data scientist — analysis, experimentation, and modelling to answer business questions. More statistics and communication, less production engineering. Range: $110k–$190k US, £50k–£95k UK.
AI product manager — defines what to build, owns evals and guardrails as product requirements, and translates model capabilities into roadmap. Range: $140k–$230k US, £65k–£120k UK.
Research scientist — invents new methods at labs and universities, usually PhD-track. Compensation at top labs is exceptional but the bar is very high.
These ranges are broad and location-dependent; a senior engineer at a frontier company can far exceed the top of these bands once equity is included.
The skills that actually get you hired
Across applied roles the common thread is software engineering plus practical model intuition. Learn Python well, understand how transformers and tokens work at a conceptual level, and get hands-on with at least one model API, a vector database, and an evaluation harness. For ML engineering, add proper training and MLOps experience. For every role, the single strongest signal is a portfolio of shipped, documented projects a hiring manager can click and try — which is why the next step after this guide is building one.