To hire AI/ML engineers in this market you have to source proactively rather than wait for applications, screen for genuine modelling and production judgement rather than buzzwords, and compete on the things money alone can’t buy — interesting problems, real data, ownership and speed. The strongest people are almost never on the job market; they’re employed, in demand and approached constantly. The win comes from a sharp, honest pitch and a fast, respectful process. Cash matters, but at scale-up budgets it’s rarely what tips the decision.
By Mark Hurren, Co-Founder, Hurren & Hope
Everyone’s chasing the same small group of people right now, so let me cut through it. Here’s how to actually land AI and ML engineers when the whole market is fishing in the same pond.
Why is hiring AI/ML engineers so hard right now?
Demand has outrun supply faster than almost any role we’ve recruited for since 2013. Every funded AI and SaaS company wants people who can take models into production, and the genuinely capable pool is small against the number of open roles. Three things make it harder:
- The title means very little. “ML engineer”, “AI engineer”, “research engineer” and “data scientist” are used interchangeably, so two people with the same title can have completely different skills.
- The best are passive. They’re shipping, well-paid, and fielding several approaches a week. A job advert will not reach them.
- The field moves monthly. Someone two years stale on tooling may still be excellent — or may have missed the shift that matters for your stack. Keyword screening gets this wrong constantly.
So you can’t hire these people the way you hire a general back-end engineer. You have to go and find them, and you have to be genuinely worth moving for.
What does a great AI/ML engineer actually look like?
Before you source anyone, get specific — because “AI engineer” covers wildly different jobs. A useful first cut:
| Profile | Core strength | You need them when |
|---|---|---|
| Applied / ML engineer | Takes models into production reliably — pipelines, training, serving, monitoring | You have a product that needs ML working, at scale, in the real world |
| Research / research engineer | Develops novel models and methods | Your edge depends on modelling innovation, not applying off-the-shelf approaches |
| AI product / LLM engineer | Builds on foundation models — RAG, agents, evaluation, prompting | You’re shipping LLM-powered features fast |
| MLOps / ML platform | The infrastructure that lets others ship ML safely | Your ML team is growing and reliability/velocity is the bottleneck |
Most scale-ups under ~200 people overwhelmingly need the applied and AI product profiles — people who ship — far more than pure research. Hiring a researcher for an applied problem (or vice versa) is one of the most expensive mismatches I see. Whichever profile, the signals of someone genuinely good tend to be production judgement (they’ve owned a model in production, not just a notebook), evaluation rigour (sceptical of their own results), pragmatism over novelty, and a real instinct for data — because the data, not the model, is usually where the wins and failures live.
Where do you actually find them?
Since the strong ones aren’t applying, sourcing is proactive and relationship-led:
- Targeted outbound to engineers doing comparable work at companies one stage ahead of you — they’ve already solved the problems you’re about to hit.
- Communities and contribution trails. Open-source ML, papers-with-code, Kaggle, model hubs and the right research communities surface real practitioners, not just well-optimised CVs.
- Referrals from your own engineers. Strong ML people know other strong ML people, and a warm intro beats any cold message.
- Specialist search when the role is urgent, confidential, or you lack the network or time — exactly the permanent technology recruitment where a consultant who has calibrated dozens of comparable hires earns their fee. See the roles we’re working on.
Don’t restrict yourself to one city if you don’t have to — a London company hiring a remote applied-ML engineer in the US is increasingly normal, though comp and competition differ sharply. Our salary benchmarks for London, New York and San Francisco map those gaps.
How should you screen and interview AI/ML engineers?
Keyword-matching CVs is where most processes go wrong. Screen for judgement and shipped work instead.
- Work, not vocabulary. Ask what they’ve actually put into production, the impact, and what failed. Specific stories beat fluent terminology every time.
- A real problem, not a brain-teaser. A focused take-home or a live discussion grounded in a problem like yours reveals far more than a generic algorithms test — and respects their time.
- Evaluation and failure. “How did you know it was working?” and “What broke, and what did you do?” separate people who’ve owned systems from people who’ve only built demos.
- Data reasoning. Probe how they think about data quality, labelling, drift and edge cases — the unglamorous things that decide whether ML works in production.
- A fast, respectful loop. Strong candidates are in multiple processes. A sprawling, slow loop loses them and signals dysfunction. We run to a roughly 23-day average time to hire for a reason.
How do you win them when you can’t outspend the giants?
A scale-up will rarely beat a frontier lab or big tech on raw cash. The good news: money is seldom the deciding factor for the people worth hiring. What moves them:
- Interesting problems and real data — often more than another £20k.
- Ownership and impact. At your size they can own a model or feature end to end and see it ship. That’s hard to find inside a giant.
- Speed and autonomy. Few layers, fast decisions, freedom to build — your structural advantage, so use it explicitly in the pitch.
- A credible technical mission and leadership that knows what it’s doing.
- A competitive, well-structured offer. You don’t need to top the market, but get the cash band right against real benchmarks and lean on equity and impact for the rest.
Where you need senior capability quickly without a permanent commitment — or to bridge a gap while you run a fuller search — contract technology recruitment can get strong AI/ML talent shipping in weeks. And where a single upfront fee would strain runway, Pay While They Stay™ spreads it over 12 monthly instalments, with payments stopping if the hire leaves.
The bottom line
Hiring AI/ML engineers in a crowded market isn’t about paying the most. It’s about being specific on the profile you need, going out to find people who’ll never apply, screening for shipped judgement over vocabulary, and moving fast with a pitch built on interesting problems and real ownership. Across 6,906 placements since 2013, with 94% still in role at 12 months, that proactive, calibrated approach is what consistently lands the hires the job-board route never reaches.
Frequently asked questions
What’s the difference between an AI engineer, an ML engineer and a data scientist?
The titles are used loosely and overlap, but broadly: ML/applied engineers take models into production and own data pipelines and serving; AI/LLM engineers build features on top of foundation models such as RAG and agents; data scientists focus on analysis, experimentation and insight. Define which job you actually need before sourcing, because the skills differ significantly.
How much do AI/ML engineers cost to hire?
Compensation varies sharply by seniority, profile and location — bands in San Francisco typically run well above London, with New York in between. Because the gaps are wide and shifting, benchmark against current market data for your specific city and level rather than a single global figure before setting a range.
Should I hire AI/ML engineers permanently or on contract?
Both have a place. Permanent hires suit core, long-term capability you want to build culture and ownership around; contractors suit urgent delivery, a specific project, or bridging a gap while you run a fuller permanent search. Many scale-ups use contract talent to start shipping quickly and hire permanently in parallel.
How do I attract AI/ML engineers when I can’t match big-tech salaries?
Compete on what large employers struggle to offer: genuinely interesting problems, real data, end-to-end ownership, speed and autonomy, and a credible technical mission. Get your cash offer into the right range against real benchmarks, then lean on impact and equity.
Hiring AI or ML talent into a fast-moving team? Talk to us about your AI hiring — we know this market and the people in it.
