How to recruit AI/ML engineers who aren’t on the market?

Your company opened an ML engineer role three months ago. The listing is up. CVs are coming in – but they’re either from juniors with one portfolio project, or from seniors whose salary expectations are out of reach. The people you’re actually looking for don’t regularly browse job boards.

 

Why standard processes don’t work

 

 

For AI and machine learning, the challenge isn’t limited to the number of available candidates. Public statistics don’t allow for a precise count of how many AI/ML engineers are working in Poland, since they cover a much broader category of ICT specialists – developers, analysts, administrators, systems engineers, and cybersecurity professionals alike.

 

Available data does, however, show the direction the market is heading. According to Eurostat 8.36% of Polish companies with at least 10 employees were using AI technologies in 2025. That’s below the EU average of 19.95% – but a clear signal that AI is moving from experimentation to real business implementation.

 

The key question isn’t just how many specialists are out there – it’s how to reach them. Many experienced AI/ML engineers aren’t actively job-hunting. They’re employed, often on B2B contracts with foreign companies that offer conditions difficult for local employers to match. A job board posting alone rarely reaches them.

 

Where passive candidates actually are?

 

 

A skilled ML engineer often leaves traces that reveal far more about their experience than a list of technologies in a CV. GitHub is one of the first places to start sourcing in this segment. Repositories with models, data tooling, contributions to libraries like TensorFlow, PyTorch, or Hugging Face, and open-source activity can all serve as concrete, verifiable competency signals. GitHub’s advanced filters let you search by programming language, topic, and recency – for example, topic:machine-learning, language:python, or pushed:>2025-01-01. This takes more effort than scanning a LinkedIn database, but gives access to people who wouldn’t appear in a standard recruitment process.

 

Kaggle is a community where data scientists and ML engineers build their skills through competitions, notebooks, and data-driven projects. Rankings like Master or Grandmaster aren’t a full replacement for production experience, but they’re a strong, externally validated signal of modeling ability, experimentation, and data skills. Many candidates also list their Kaggle rankings directly on LinkedIn.

 

It’s also worth looking beyond online platforms. Conferences and communities – MLOps Community, PyData, Data Science Warsaw, Krakow Data Science – bring together people who engage with the subject because they work professionally in AI and machine learning, not because they’re actively looking for their next role.

 

The challenge is that operating effectively in these channels requires technological context. A recruiter doesn’t need to build models – but they should be able to tell a demo project from real production experience, and hold a conversation that doesn’t begin and end with a list of tools..

 

What sets a good AI/ML hiring approach apart

 

AI/ML candidates receive recruiter messages constantly. Most of them sound nearly identical: “Hi, I have an exciting opportunity for someone with your profile.”

 

Personalizing the name isn’t enough anymore. Good sourcing in this segment starts with genuinely engaging with the candidate’s work – their repositories, articles, talks, or open-source contributions. If you’re reaching out to someone who contributed to a language model toolchain or maintains an actively-used repository, your first message should show that you understand what they do and why their experience is relevant to the specific problem your company is trying to solve.

 

This takes more time per contact than mass outreach. But it meaningfully increases the chance of starting a conversation with a candidate who would otherwise ignore another generic message.

 

The second challenge is an offer that doesn’t match the candidate’s context. Someone working on an interesting product or running a contract for a foreign company won’t switch jobs just because a new role is “in the AI space.” They need to see real value: an ambitious problem to solve, influence over product and technical decisions, access to good data, an interesting stack, or the chance to build something from scratch.

 

That’s why a recruitment conversation should start not with presenting the offer, but with understanding what would represent a meaningful change for that particular person.

 

The competency assessment problem

 

What sets the AI/ML market apart from traditional IT recruitment is how difficult it is to verify skills.

 

Familiarity with PyTorch, Scikit-learn, or TensorFlow can be claimed in a CV. What’s far harder to assess is whether a candidate can build a production pipeline, evaluate data quality, understand model drift, and translate results into business decisions.

 

Google’s materials on production ML systems note that creating a model is only the beginning. In practice, you need to monitor data, features, prediction quality, differences between training and production environments, pipeline stability, and real-world model behavior after deployment. Similarly, the NIST AI Risk Management Framework emphasizes continuous testing, measurement, validation, and risk management across the entire AI system lifecycle.

 

In practice, this means that CV assessment is less reliable here than in many other technical roles. Companies that recruit AI/ML talent effectively build selection processes based on concrete tasks and case studies – not just lists of technologies in a CV.

 

This requires the hiring manager involved from the start – and a recruiter who understands what to ask.

 

In an interview, it’s worth checking:

 

• what business problem the candidate worked on;
• what data was available and what its limitations were;
• how they measured model quality;
• what didn’t work in the first version of the solution;
• what the deployment process looked like;
• whether and how the model was monitored post-launch;
• what technical decisions the candidate made independently.

 

 

This gives far more information than asking about familiarity with the next framework.

 

Time and cost – What to expect

 

There’s no single reliable publicly available statistic showing average time-to-fill for senior AI/ML engineers in Poland. Data from recruitment reports varies by methodology, company group, and role definition – making it difficult to treat as a universal benchmark.

 

In practice, process duration depends on seniority level, specialization, collaboration model, product appeal, budget, speed of decision-making, and whether the organization actually knows what it’s looking for.

For any company, more important than a market average is observing your own process. If after several weeks of active sourcing there are no meaningful conversations, the problem isn’t always a shortage of candidates. It often means the role is too broadly defined, the offer doesn’t reflect market realities, or the selection criteria are unclear.

An unfilled ML engineer role for several months isn’t just a recruitment cost. It also means delayed product delivery, missing automation, slower data analysis, or the risk that a project won’t reach production on schedule.

HR Contact sees this pattern regularly across EMEA, USA, and LATAM markets. Companies that close processes faster combine two things: active sourcing in non-obvious channels and clearly defined competency criteria from the start. Without the latter, even the best candidate list leads to extended decision-making.

 

When to bring in an external partner?

 

AI/ML is one of the segments where the difference between a recruiter who knows the market and one who doesn’t is greatest – not because AI is hard to understand, but because it’s easy to mistake someone who knows the vocabulary for someone who can actually evaluate a project.

 

External support is worth considering when a role has been open for more than 8 weeks without concrete candidates in the pipeline, when your internal HR team doesn’t have access to the sourcing channels described above, or when the company is planning to build an entire AI team – not fill one role.

 

 

Sourcing on demand allows you to launch an active pipeline without committing to a full recruitment process. Embedded recruitment makes sense when a company hires regularly and needs someone who genuinely understands the technical context – not just processes CVs.

 

FAQ – Questions About AI/ML Engineer Recruitment

 

How long does it take to recruit a senior AI/ML engineer?

 

There’s no single reliable average for the Polish market. A process can close quickly if the company has a well-defined problem, a realistic budget, an efficient decision-making process, and actively reaches out to candidates. It can also take months if the organization relies solely on job postings, has no clear role definition, or changes expectations after every interview.

What matters most isn’t counting the weeks – it’s detecting the problem early. If the first rounds of interviews aren’t surfacing the right profiles, it’s better to adjust your sourcing strategy or refine the role than to wait for the right CV to arrive by chance.

 

How do you verify whether an AI/ML candidate actually knows what they claim on their CV?

 

CVs are less reliable in AI/ML than in traditional software development. It’s worth verifying GitHub repositories, public projects, open-source contributions, Kaggle activity, articles, talks, or descriptions of specific deployments.

 

he most important thing, however, is a conversation about a real project. A candidate should be able to explain not just which model they used, but where the data came from, how they measured quality, what problems arose, and what they did after deployment when the model started running on real-world data.

 

Summary

 

AI/ML engineer recruitment doesn’t work like recruiting standard developers. The market is narrow, candidates are often passive, and assessing competencies requires a different approach than verifying a tech stack.

 

Companies that wait solely for CV flow from job boards limit their access to a large portion of the market. An effective approach starts with sourcing in places where candidates actually leave traces of their work: GitHub, Kaggle, technical communities, and conferences. It requires personalized outreach, understanding of each candidate’s motivations, and a competency assessment process that holds up against well-written CVs.

 

If you have an open AI/ML role that hasn’t been filled for several months – talk to HR Contact.

 

---