Skip to main content
Career / AI Jobs
Skills for an AI Career
AI careers aren't one thing — they range from deep ML research to product management to AI ethics. Here's what each path actually requires, with salary context and how to get there.
The market reality
97M
New AI-adjacent jobs the World Economic Forum projects by 2025 — but concentrated in specific skills, not evenly distributed [WEF Future of Jobs 2023]
$185K
Median total compensation for ML engineers at major US tech companies in 2025, according to Levels.fyi — up 23% from 2022 [Levels.fyi]
3x
More AI/ML job postings than qualified candidates, per LinkedIn — a persistent skills gap that shows no sign of closing despite increased training programmes [LinkedIn Jobs Report]
AI career paths and salaries (2026)
| Role | Core skills | UK salary range | US salary range |
|---|---|---|---|
| ML Engineer | Python, PyTorch/TensorFlow, model training, MLOps, cloud (AWS/GCP) | £65K–£120K | $140K–$250K+ |
| Data Scientist | Python, SQL, statistics, pandas/sklearn, ML fundamentals, visualisation | £50K–£95K | $110K–$180K |
| AI Product Manager | Generalist PM skills + AI/ML literacy, prompt design, evaluation, stakeholder management | £60K–£110K | $130K–$220K |
| AI Engineer (LLM) | Python, LLM APIs, RAG, vector databases, agents, prompt engineering, evals | £55K–£105K | $130K–$220K |
| Prompt Engineer | Deep knowledge of LLM behaviour, systematic evaluation, Python basics | £45K–£85K | $100K–$160K |
| AI Ethics / Policy | Policy/legal background, AI literacy, risk assessment, stakeholder communication | £50K–£90K | $100K–$170K |
| MLOps Engineer | DevOps, cloud, model deployment, monitoring, CI/CD for ML pipelines | £60K–£105K | $130K–$200K |
| AI Research Scientist | PhD-level maths, deep learning theory, research publication, Python | £80K–£150K+ | $180K–$350K+ |
Skills breakdown by role type
ML Engineering / Data Science
- MustPython — pandas, numpy, scikit-learn fluency
- MustSQL — data querying and manipulation at scale
- MustDeep learning framework — PyTorch or TensorFlow
- MustStatistics and probability foundations
- NiceCloud ML platforms — AWS SageMaker, Vertex AI, Azure ML
- NiceMLOps — experiment tracking (MLflow), model serving (BentoML)
AI Engineering (LLM / Application)
- MustPython + LLM API integration (OpenAI, Anthropic, Hugging Face)
- MustRAG systems — vector databases (Pinecone, Chroma, pgvector)
- MustPrompt engineering and systematic evaluation
- MustLLM application frameworks — LangChain or LlamaIndex
- NiceFine-tuning with LoRA/QLoRA via Hugging Face
- NiceAI agents — tool use, multi-agent frameworks
AI Product Management
- MustAI/ML literacy — understanding capabilities and limitations
- MustEvaluation design — defining success metrics for AI features
- MustStakeholder communication about AI uncertainty and limitations
- MustStandard PM skills — discovery, prioritisation, roadmapping
- NiceSQL for data analysis and experiment interpretation
- NicePrompt design — enough to specify feature requirements clearly
AI Ethics and Policy
- MustAI/ML literacy — enough to evaluate risks and capabilities claims
- MustRegulatory awareness — EU AI Act, sector-specific rules
- MustRisk assessment frameworks — bias auditing, impact evaluation
- MustDomain expertise — law, policy, healthcare, or finance background
- NicePython basics — enough to audit model outputs independently
- NiceFairness metrics — understanding demographic parity, equalised odds
The hiring manager's perspective
The most in-demand AI engineers in 2026 are those who can bridge model capability and production deployment — people who understand LLM behaviour well enough to design reliable systems, not just call APIs. The second most sought-after profile is people with deep domain expertise (medicine, law, finance) plus AI literacy — because AI applied to complex domains needs someone who can evaluate quality, not just measure benchmark scores. Pure "AI generalist" roles are increasingly hard to fill with 6-week bootcamp graduates — employers want either deep technical ML skills or deep domain expertise plus AI application skills. Building a portfolio of working projects — even personal ones — remains the most effective way to break into AI roles.
FAQ
Do I need a degree to work in AI?
For ML research roles at major labs: almost universally yes — a PhD or strong master's is expected. For engineering and application roles: less so. Many of the strongest AI engineers are self-taught or came through CS bootcamps. What matters is a demonstrable portfolio of work: GitHub projects, open source contributions, Kaggle competition performance. For product, policy, and adjacent roles: a relevant domain degree (law, medicine, business) is often more valuable than a computer science one, combined with demonstrated AI literacy.
What's the fastest path into an AI role from a non-technical background?
AI Product Manager is the most accessible entry point for people with domain expertise but limited technical background — it leverages existing skills (strategy, communication, product sense) plus AI literacy rather than requiring engineering depth. AI ethics and policy roles are a good fit for lawyers, policy professionals, and ethicists. For people willing to learn Python, AI engineering via the LLM application path (APIs, RAG, agents) requires less mathematical background than traditional ML and has strong job market demand. The fastest path is always to combine your existing expertise with AI literacy, rather than trying to become a pure AI specialist from scratch.
Get AI insights every week
The AI Briefing covers what actually matters in AI — no hype, no jargon, just what you need to stay ahead.
Subscribe free