We've been working in the AI space for the past couple of years, and in that time we've seen a load of new job titles, each with a slightly different meaning depending on who you ask.

Recruitment at its basic is pattern recognition; so it matters that you and your client are using the same vocabulary. If you're both saying "Senior AI Engineer" but picturing different things, the pattern you're matching against is different from the start.

So we put together this guide internally to make sure our consultants are aligned with clients from day one. It's designed to simplify a complex market.

Here's how we see the AI landscape right now:

AI Researcher - generates new capabilities. Publishes. Works in theory and experimentation. Academic lineage matters because the community is small and credentialling is real. Rare, expensive, not looking, and largely unimpressed by equity decks.

ML Engineer - trains, fine-tunes, and evaluates models. Bridges research and production. Knows PyTorch the way a senior backend engineer knows their database. The most legible role in the stack. Sits across the research-applied boundary depending on the company — at a big lab they're close to research; at a startup they're shipping product.

LLM Engineer - specialises in language model behaviour: prompting, RLHF, evals, alignment-adjacent work. A role that barely existed before GPT-3. Still finding its shape.

Applied AI Researcher - the hybrid most companies say they want and struggle to find. Research-trained, but building real systems against real deadlines. Tends to come from applied research teams at the big labs or from the handful of academic groups that also ship products. You'll probably hear "high agency" somewhere in the job spec.

Alignment Researcher - studies how models and AI systems behave safely and predictably. Splits into two distinct flavours: model-level work on values, objectives, and training behaviour; and systems-level work on monitoring, evals frameworks, and interpretability infrastructure. The titles look identical from the outside.

Systems Researcher - does novel research on the infrastructure underpinning AI: retrieval algorithms, vector indexing, distributed systems at scale. Lives mostly in academia and big lab research divisions. Rarely hired commercially but the work feeds into everything.

Evals Researcher - designs and studies how to measure AI system performance. Research-grade work, but on the scaffolding rather than the model itself. An emerging specialism as the question of how to evaluate AI gets harder.

AI Engineer - builds with models rather than on them. API integration, pipelines, orchestration. Closer to software engineering than ML. The fastest-growing category and the most oversaturated job title in tech right now. Requires real due diligence — this title covers a huge range.

RAG Engineer - a sub-specialism of AI Engineer focused on retrieval: chunking, embeddings, vector databases, hybrid search, reranking. More of a phase people pass through than a career identity, though that's starting to change as retrieval architecture gets genuinely complex.

MLOps / AI Platform Engineer - infrastructure: model serving, latency, cost, monitoring, drift. Common in enterprise; in smaller companies, someone else absorbs this as a bolt-on. Often invisible until something breaks.

DevEx Engineer - builds the tooling, SDKs, and internal platforms that let other engineers work with AI systems effectively. Closer to MLOps than ML. Often undervalued until the developer experience is bad enough that everyone notices.

Data Engineer - feeds everything upstream. Not AI-specific but increasingly the bottleneck, as data quality is often what actually determines whether your AI works.

Software Engineer (AI-aware) - writes product code, integrates AI features, can't ignore the stack but isn't a specialist in it. The largest group by headcount and growing fast.

How to read the market

These jobs overlap in different ways, it's useful to view them across two axes.

The first is research vs. applied. Are they generating new capabilities or deploying existing ones? Research people are asking questions the field hasn't answered & applied people are turning answers into products.

The second is model-level vs. system-level. Are they working inside the model - training, evals, behaviour, or around it, APIs, pipelines, retrieval, orchestration?

Put those together and you get four quadrants. Most roles fall cleanly into one, though ML Engineer doesn't fall neatly into one camp, sitting closer to research at the big labs and closer to applied at startups.

We've put the job titles in the quadrants where we see them, but it's important to know that in reality you could see these titles in different boxes. AI Engineer, for example, is so generic it could sit in any of them.

The second useful frame is the split between builder and researcher. The frontier labs tend to hire the researchers, and they tend to publish their papers online, so it's reasonably easy to work out what someone specialises in, though it's time-consuming. Builders you tend to see in the YCombinator-style startups.

Someone who is an elite-level researcher and builder is extremely rare.

It's also worth mentioning the title "Technical Member of Staff", an old Bell Labs term deliberately revived to keep roles open-ended, emphasise collective expertise, and reduce hierarchies. In an AI lab, it's a reasonable bet that someone with this title sits in the builder/researcher spot, but not guaranteed.

Before you start working with anyone hiring in the AI space, align on two things:

1. Where does the work sit on the research-applied axis, and where does it sit on the model-system axis?
2. How much of a builder vs. researcher do they need?

If you're all aligned on what that means, you've won half the battle.

A lot of briefs aren't explicit on either as it's hard to know what you need when the field is moving this fast and the titles mean so many different things. But fixing the brief is our job, and it's worth doing before you spend six weeks finding out you've been looking for the wrong person.

If you want someone in the top range, top university, industry experience, in the office and can't pay a minimum of £150k, we'd want to have that conversation before we start.