Cómo contratar a tu primer ingeniero de IA: Una guía para startups

For many startups, the first AI engineer is one of the most important hires they will make. This person will shape how AI works in your product, which tools and models you build on, and often who you hire next. Get it right and you move faster than competitors. Get it wrong and you can spend months, and a lot of money, on prototypes that never reach customers.

It is also one of the hardest hires to make. “AI engineer” means different things to different companies, the strongest candidates have plenty of options, and many founders aren’t sure how to assess the skills they need. This guide walks through how to hire your first AI engineer, from defining the role to making an offer.

1. Decide what you actually need the AI engineer to do

Before you write a job description, be specific about the problem. Most first AI hires at startups fall into one of three types of work.

• Building AI features on top of existing models. Using large language models and other foundation models through APIs, connecting them to your data, and shipping features customers use. This is the most common need today and is often called applied AI engineering.
• Training or fine-tuning your own models. Working with your own data to train, fine-tune and evaluate models. This needs deeper machine learning experience and usually more infrastructure.
• Research. Developing new methods or models. Very few early-stage startups need this as a first hire.

Write down what you want the person to have shipped after three, six and twelve months. If the answer is “an AI assistant in our product that customers use every day”, you need a strong applied AI engineer, not a research scientist.

AI engineer, ML engineer or data scientist?

These titles overlap and companies use them loosely, which confuses candidates as much as employers. As a rough guide:

• An AI engineer builds products and features using AI models, usually with strong software engineering skills.
• A machine learning engineer focuses on training, deploying and running models in production, including the data pipelines and infrastructure around them.
• A data scientist focuses on analysis, experimentation and statistical modelling to answer business questions.

For most startups building AI into their product, the first hire should be an engineer who can ship, not an analyst. We look at some newer roles in more detail in 4 roles que están dando forma a la próxima generación de equipos de IA.

2. Hire for software engineering first, AI second

The most common mistake we see is hiring someone who is impressive in AI but can’t build reliable software. Your first AI engineer will probably work without a large platform team around them. They need to write production-quality code, design systems that handle real traffic, and work with your existing engineers.

Look for evidence of:

• AI features or products they have shipped to real users, not only demos or notebooks.
• Strong fundamentals in your main language and stack.
• Experience evaluating model output and measuring quality, not just making something that works once.
• Awareness of cost, speed and reliability, because model calls can get expensive and slow.
• Good judgement about data privacy and security.

3. Decide seniority before you set the budget

A first AI engineer usually needs to work with a lot of independence, so most startups should aim for a senior or experienced mid-level engineer rather than a junior. A junior engineer can do excellent work with guidance, but someone has to provide that guidance, and at this stage there often isn’t anyone.

Salaries vary widely by location and seniority. In Barcelona, for example, Blu Selection’s 2026 salary guide puts junior AI engineers (one to three years’ experience) at around €40,000 to €55,000 and senior AI engineers (eight years or more) at around €85,000 to €110,000 base salary. London and other major hubs are typically higher. Check current market data for your location before you go out to candidates, because offers that are too low are one of the main reasons AI hires fall through late in the process.

For a wider view of costs in one European market, see the real cost of hiring tech talent in Spain.

4. Write a job description candidates want to read

Strong AI engineers receive a lot of messages from recruiters. A vague advert asking for “AI/ML experience” with a long list of tools won’t stand out. Instead:

• Explain the problem they will solve and why it matters to your customers.
• Be honest about what exists today: your data, your stack, and what has already been tried.
• Say who they will work with and who they will report to.
• Describe what success looks like in the first year.
• Include the salary range if you can. It saves time on both sides.

5. Assess real skills, not just theory

Traditional technical interviews often miss what matters in AI engineering. A good process for a first AI hire usually includes:

1. A conversation about past work. Ask them to explain something they built: the problem, the choices they made, what went wrong and how they measured success.
2. A practical exercise close to your real work. For example, a short task to build a small feature using a model, followed by a discussion of how they would evaluate it, improve it and run it in production. Keep take-home tasks short and respect candidates’ time.
3. A system design discussion. How would they design the AI part of your product? Listen for trade-offs on cost, speed, accuracy and data.
4. A conversation with the founders. Your first AI engineer will influence product direction, so they need to understand and believe in what you’re building.

If nobody on your team can judge AI skills in depth, consider bringing in an experienced engineer or adviser for the technical interview, or working with a recruitment partner who specialises in AI roles.

6. Move quickly and sell the opportunity

Good AI engineers are often in several processes at once. Long gaps between interview stages are one of the easiest ways to lose them. Aim to complete the process within two to three weeks of the first interview, give feedback within a day or two, and keep candidates updated even when there’s nothing new to say.

Remember that candidates are assessing you too. Be ready to talk about your funding, your roadmap, the data they will have access to, and how much ownership they will have. For a first AI hire, the chance to shape a company’s AI work from the start is often the strongest selling point you have.

7. Plan for their first 90 days

Hiring is only half the job. Before your new engineer starts, make sure they will have access to the data, tools and budget they need, a clear first project, and regular time with the founders. A strong start helps them deliver early and makes it far more likely they stay.

Common mistakes to avoid

• Hiring a researcher when you need someone to ship product.
• Writing a “unicorn” job description that asks for every AI skill at once.
• Underestimating salary expectations in a competitive market.
• Running a slow process with too many interview stages.
• Hiring before you have the data or a clear problem for them to work on.

Key takeaways

• Define the problem and what you want shipped before you define the role.
• For most startups, the first AI hire should be a strong software engineer who builds with AI.
• Aim for senior or experienced mid-level, and benchmark salary against current local data.
• Use practical assessments that reflect your real work, and keep the process fast.

Need help hiring your first AI engineer?

Global M helps startups and scale-ups hire engineers and AI specialists across the UK, Spain and beyond. With Contratación basada en el éxito, we run the search for key roles like your first AI engineer and you only pay when you hire. If you’re building a whole AI or engineering team, our Equipo de contratación integrado gives you dedicated Talent Partners for one monthly fee.

Book a free 20-minute call to talk about the role you’re hiring for.

Further reading: What makes a great first engineering hire? 5 traits you shouldn’t overlook.