The AI hiring pillar
Hire AI engineers: the complete 2026 guide.
"AI engineer" is not one job — it is at least seven. This guide maps every AI engineering role we place, how the 5-day shortlist process works, what it costs against the 22–33% agency market, and how the UK and German markets differ.
Scope any AI role on a 30-min call and we deliver a vetted 3-candidate shortlist in 5 business days. 15% success fee, £0 upfront, 90-day replacement guarantee.
Who this is for
Who this guide is for
This page is written for founders and engineering leaders at Series A–C product scale-ups — in the UK, Germany and DACH, the Netherlands, the Nordics, and the US — who need to make an AI engineering hire and are staring at a job title that has stopped meaning anything. "AI engineer" in 2026 is an umbrella over at least seven distinct jobs, each with its own skill profile, its own failure modes, and its own imposter problem. Writing one JD that mixes three of them is the most common way an AI role stays open for a quarter.
It is also written for the buyer who has already tried the alternatives. If you have run the generic "hire three, accept one" funnel, you know it ties up 40–60 hours of senior engineer interview time per placement — the most expensive part of the whole exercise. If you have briefed a classical contingency agency, you know most cannot tell a candidate who has operated an inference stack from one who has read about it, because their screen stops at keyword matching.
Recruo is an AI-augmented recruitment agency: we source senior AI engineers across Central and Eastern Europe, screen them with an adaptive AI technical interview plus human recruiter review, and deliver a 3-candidate shortlist in 5 business days at a 15% success fee. Each of the seven roles below links to a dedicated deep-dive page with hiring-difficulty benchmarks from our own funnel, an anonymised placed-candidate profile, and role-specific FAQs. This page is the map; those pages are the territory.
The seven roles
The 7 AI engineering roles, and when to hire each
These are the seven AI engineering roles we shortlist for, in the order most scale-ups need them. The one-line rule that resolves 80% of scoping calls: at 1–3 AI engineers hire the generalist, above 3 specialise into the role that matches your product moat, and add the platform engineer once the team passes eight. Each card links to the full role guide.
AI/ML engineer (generalist)
The senior generalist who owns the whole AI surface at a Series A–B scale-up: fine-tuning small task-specific models, wiring OpenAI or Anthropic calls into product flows, standing up pgvector for semantic search, and writing the throwaway eval harness that decides which model earns its cost. The single point of AI literacy until hiring catches up.
Hire when: Your AI team is 0–3 engineers and nobody owns AI end-to-end yet. This is the default first AI hire.
LLM engineer
Owns the end-to-end cost, quality, and latency of the LLM-backed product surface — prompt design and output evaluation, retrieval and context shaping, and the serving-side optimisation that turns a 900ms p95 into a 180ms one. The most-abused job title in tech: most self-titled candidates have shipped demos, not systems under load.
Hire when: A text-generation feature is core to the product and its latency, cost, or quality has a named budget.
RAG engineer
Owns the retrieval and context-assembly layer that feeds a production LLM: chunking strategy, hybrid BM25+dense retrieval, reranking, context compression, and the eval harness that tells you whether any of it is helping. Distinct from LLM engineering — the generator is a commodity; the retrieval layer is where the quality signal lives.
Hire when: Retrieval over a proprietary corpus — docs, tickets, contracts, code — is the product moat.
AI evals engineer
Ships the evaluation loop: golden-reference datasets versioned like code, automated eval harnesses wired into CI, LLM-as-judge calibrated against human raters, and the adversarial and safety suites you need if your product falls under Annex III of the EU AI Act. The role standing between a prompt change and a silent quality regression.
Hire when: You ship LLM features regularly but nobody can say whether last week’s release made quality better or worse.
Agentic AI developer
Owns the orchestration and control-flow layer: turning a single-turn completion into a multi-step, tool-using, self-correcting agent that survives contact with a production user. Fluent in LangGraph, first-party agent SDKs, and MCP — and honest enough to talk you out of an open-loop agent when a deterministic workflow ships faster.
Hire when: You have an agentic workload in production or a funded 90-day pilot, and no dedicated owner for it.
ML platform engineer
Owns the AI infrastructure layer your product teams consume: inference serving with vLLM or SGLang, GPU orchestration on Kubernetes and Ray, the model registry, and the cost dashboards that stop an inference bill becoming a board-level incident. Reasons natively in GPU utilisation, tokens-per-GPU-hour, and cost-per-million-tokens-served.
Hire when: Your AI team has passed ~8 engineers, or you run models on a mix of APIs and self-hosted open weights.
Generative AI engineer
The broadest of the modern AI product roles: owns whatever modality the product needs — text, image, audio, and increasingly all three at once. Part machine learning, part product engineering, part procurement: knowing when to call a frontier API, when to fine-tune an open model, and when generative AI is the wrong tool entirely.
Hire when: The product touches multiple generation modalities, or nobody has decided yet which models to standardise on. A strong second or third AI hire.
The boundaries matter more than the titles. The LLM engineer owns model selection, prompting, and serving; the RAG engineer owns retrieval and context assembly; the agentic developer owns orchestration and control flow; the evals engineer owns the loop that tells all three whether their work is helping. If a single JD asks for all four, you are not hiring a unicorn — you are hiring a generalist, and you should scope (and pay) it that way.
How it works
From mandate to shortlist in 5 business days
The same 6-step flow runs for every role on this page. AI does roughly 80% of the screening work; a human recruiter reviews and signs off on every shortlist before it reaches you — that human-in-the-loop step is a GDPR and EU AI Act design requirement, not a marketing line. Individual-contributor shortlists land in 5 business days; VP and Head-of-AI searches take 10–14.
Intake call
30-min scoping. We capture the JD, must-haves, culture, budget, and timeline — and tell you honestly which of the seven roles you actually need.
Sourcing
LinkedIn, DOU, Djinni, Just Join IT, Bulldogjob, open-source contributor graphs, and our private CEE network of AI engineers.
AI CV screening
Every CV scored against the JD. Only 70+ scores move forward — for LLM roles that filters out roughly 86% of inbound CVs.
Soft-skills interview
20–30 min recruiter call: motivation, English level, salary expectations, availability.
AI technical interview
Adaptive 12-minute technical interview on video with anti-cheat checks, probing production signals — not framework trivia.
Shortlist delivery
3–5 candidates with CV analysis, the AI interview report, and recruiter notes. A human recruiter signs off on every shortlist before it reaches you.
What lands in your inbox is not a stack of CVs: each candidate arrives with a CV analysis, the full AI interview report, and recruiter notes, so your team walks into the first interview already knowing where to dig. If the first shortlist does not produce a hire, we re-run sourcing at no extra cost. Every placement carries a 90-day replacement guarantee.
What it costs
Pricing: 15% success fee vs the 22–33% agency market
The classical agency market prices AI engineering hires the same way it prices everything else: UK contingency agencies run around 22% of first-year salary, German Personalberatungen 25–30%, US contingency around 25%, and retained search up to 33%. Recruo charges a 15% success fee — on an €85K senior LLM engineer that is €12,750, paid only when the candidate signs. Nothing upfront, no retainer, no exclusivity demand. Overall that works out roughly 32% cheaper than classical agencies. Full breakdown and a savings calculator are on the pricing page.
For junior and mid-level roles there is a second model: a $8,000 fixed fee per placement, for teams that want predictable spend regardless of salary. And for a first engagement, the First Hire Pilot takes 25% off the Standard fee — an 11.25% effective rate on your first placement — so you can validate the flow before committing to more.
The larger saving is usually on the salary itself. Because we source across Central and Eastern Europe, the underlying compensation runs 40–55% below UK and German local rates: a senior engineer at €60–75K in Warsaw or Kyiv is the equivalent of £100–130K in London — same quality bar, 7–9 hours of daily time-zone overlap, English as the default working language. That delta is a local-market gap, not a quality gap, and the Eastern Europe hiring guide breaks it down country by country.
One honest caveat on comparisons: a success fee and a marked-up hourly rate are different products. If you need short-term contract capacity rather than a long-term hire, a Toptal- or Proxify-style model can be the right call — we lay out the trade-offs on our vs Toptal and vs Proxify pages.
Market notes
Hiring AI engineers in the UK and Germany
United Kingdom. The compliance question UK clients ask first is IR35. Every engineer we shortlist operates their own registered business in CEE — multiple clients, own tools, own hours — which HMRC treats as outside IR35 by default; for inside-IR35 edge cases we arrange an Employer-of-Record in the engineer’s home country so your UK entity carries zero payroll exposure. We keep UK-specific editions of the role guides with local salary banding and compliance notes: LLM engineers (UK), RAG engineers (UK), AI evals engineers (UK), agentic AI developers (UK), ML platform engineers (UK), and AI/ML engineers (UK) — plus the broader UK recruitment agency page.
Germany & DACH. The German equivalent is Scheinselbstständigkeit risk under §7 SGB IV. The same structure answers it: engineers contract from a registered business outside Germany (JDG in Poland, ФОП in Ukraine, PFA in Romania), and where DRV risk exists we default to EOR employment, keeping your GmbH with a clean audit path. We sign an AVV/DPA on every engagement and keep candidate data in EU regions. Against the local 25–30% Personalberatung norm, the 15% fee compounds with the salary delta. German-market editions: LLM engineers (DE), RAG engineers (DE), AI evals engineers (DE), agentic AI developers (DE), ML platform engineers (DE), and AI/ML engineers (DE) — plus the Germany recruitment page.
FAQ
Frequently asked questions
Get started
Get your 3-candidate shortlist
Scope one open AI engineering role and get a vetted shortlist in 5 business days. £0 upfront, 90-day replacement guarantee.
No upfront fees
Pay only on placement
90-day guarantee
Free re-search if hire leaves
EU AI Act ready
Compliant by design
