Top 10 Companies Hiring AI Engineers in Belgium in 2026

Table of Contents

• Introduction
• Proximus
• Collibra
• imec
• BNP Paribas Fortis
• UCB
• KBC
• Google
• Microsoft
• Showpad
• Solvay
• How to choose your Belgian AI employer
• Frequently Asked Questions

Proximus

What you actually build

From a desk inside the Brussels Ring, Proximus feels less like a legacy telco and more like a national infrastructure lab. Through its Proximus ADA centre of excellence, the group is running over 150 active AI projects, ranging from 5G optimisation to fraud analytics. In their own expert talks, they describe hard-won lessons from “over 150 AI projects” across network, customer and enterprise use cases, emphasising that these are not toy models but systems wired into the live network (Proximus NXT lessons from AI projects).

Much of this sits at the intersection of AI and security. ADA is one of Belgium’s first units built explicitly around that pairing, handling:

• Real-time anomaly detection on telecom traffic for cybersecurity
• Churn and upsell prediction on millions of customer journeys
• 5G and fibre network optimisation under strict SLAs

How engineering actually works

The stack is unapologetically industrial: cloud-first on Microsoft Azure, heavy Python and PyTorch, plus dedicated MLOps platforms. Proximus repeatedly stresses that robust data foundations make or break projects; one of their own blogs flatly concludes that “an AI project is a data project” (AI-as-data-project overview).

“An AI project is a data project.” - Proximus NXT expert talks, Proximus

Why it matters in Belgium

For a Belgian ML engineer, the appeal is threefold: nationwide impact (models that touch almost every SIM and fibre line in the country), proximity to EU digital policy in Brussels, and a strong role in the DigitAll digital inclusion ecosystem alongside banks and public bodies. Senior AI profiles typically land in the upper half of Belgium’s €75k-€115k senior ML band, with the usual Belgian corporate perks layered on top.

If you want to work on petabyte-scale logs, real-time constraints and policy-relevant cybersecurity - without leaving Brussels - Proximus is one of the most substantive “pralines” in the box.

Collibra

What you actually build

Collibra is the Brussels-born data unicorn behind one of the world’s leading data intelligence platforms. As an AI engineer here, you work on the “plumbing” that makes analytics and GenAI safe to use: NLP for automated data cataloguing, metadata extraction and entity resolution across thousands of systems, and ML that scores data quality and flags policy violations in real time. In global rankings of Belgian AI startups, Collibra regularly appears as the flagship scale-up for data governance and cataloguing platforms (Seedtable’s overview of Belgian AI startups).

How the AI gets built

Under the hood, the platform runs as multi-tenant SaaS on AWS, using SageMaker, Kubernetes and a Python/PyTorch-heavy stack. Rather than building one-off bespoke models, you focus on capabilities that must work across hundreds of enterprises: semantic search over metadata, entity linking between noisy schemas, and policy engines that can answer “who can see what?” for auditors in seconds.

• Designing ML services that scale across customers and regions
• Embedding models directly into workflow UIs and APIs
• Hardening everything for privacy, latency and uptime

Why it fits the Belgian ecosystem

Collibra sits in a sweet spot for Belgium: HQ in Brussels, strong links to US and EU clients, and a remote-first culture that still values time in local hubs. Its recent Series G round of £184.3M was explicitly framed around growing its data privacy and governance capabilities, reflecting how central these topics are as the EU’s AI Act takes shape (Tech Funding News coverage of Collibra’s funding).

If you want your ML skills to travel well between Brussels, Amsterdam and other EU capitals - and you enjoy building the foundational layer that keeps everyone else’s AI honest - Collibra is one of the most strategically placed employers in the Belgian ecosystem.

imec

What you actually build

In Leuven, imec is what happens when a chip lab and an AI lab share the same espresso machine. Officially it’s a world-leading centre for nanoelectronics and digital technologies; day to day, that means AI engineers working on ML performance engineering for next-generation chips, hardware-aware optimisation with tools like JAX and C++, and models that shape semiconductor design and advanced materials research.

Instead of optimising click-through rates, you are tuning training and inference on novel architectures, contributing to high-performance computing programmes such as EuroHPC DARE, and pushing AI into domains like thin-film magnetic materials, medical devices and automotive systems through imec’s industrial partnerships.

How engineering feels from the inside

The culture is unapologetically research-heavy. imec sits physically and intellectually inside the KU Leuven ecosystem, and many engineers arrive via PhDs in ML, control or computer architecture. Your week can span:

• Co-designing algorithms and chips with hardware teams
• Profiling and optimising end-to-end ML pipelines on experimental hardware
• Publishing results and co-supervising theses with university labs

Unlike classic web product work, success here is measured in benchmarks, conference papers and design wins with global semiconductor partners as much as in internal KPIs.

Why it matters in Belgium

imec anchors the Leuven-Ghent deep-tech corridor and regularly appears among the most attractive employers in Belgian ML job overviews, which highlight it as a unique chance to work at the software-hardware frontier within Belgium’s borders (Belgian ML engineer job listings that feature imec). Positions often sit at or above the upper end of the Belgian senior ML salary range, reflecting the international competition for this niche skill set.

If your ideal “praline” is less SaaS dashboard and more “between lab and fab”, imec is the flagship option in the Belgian box: a place to work on chips, physics and AI in the same commit.

BNP Paribas Fortis

What you actually build

At BNP Paribas Fortis, “AI in banking” means large, long-lived systems rather than isolated POCs. AI and ML engineers work on production platforms for fraud detection, anti-money laundering, automated risk scoring and credit decisioning, plus personalisation in the mobile and web banking channels. These systems underpin one of Belgium’s most mature ML infrastructures in finance, with roles explicitly scoped around industrialising models rather than just experimenting in notebooks.

Typical responsibilities in senior roles include designing and deploying services for transaction monitoring, customer profiling and risk modelling, as reflected in the bank’s own Senior AI/Machine Learning Engineer job descriptions.

How engineering runs

The tech stack leans heavily on Python, solid object-oriented design and MLOps with GitLab. You work in multidisciplinary squads with data engineers, risk officers and product teams, building pipelines that can stand up to audits from internal model-risk committees and external regulators.

• Implementing ML models into high-availability services
• Ensuring full traceability and explainability for every decision
• Managing versioning, testing and rollback for regulated workloads

Why it matters in Brussels

BNP Paribas Fortis sits right in Brussels’ finance-EU corridor and is a key member of the national DigitAll digital inclusion alliance, which brings together banks, telcos and public bodies to make digital services more accessible across Belgium (coverage of the DigitAll alliance launch).

For senior ML engineers, compensation typically lands around €85,000-€105,000 plus a classic banking package: pension contributions, extensive insurance and potential performance bonuses. If your goal is to learn how to run governed, explainable ML at European banking scale - while staying close to both the EU institutions and the Belgian tech scene - BNP Paribas Fortis is one of the most strategic choices in the box.

UCB

What you actually build

At UCB, AI engineering sits right inside the drug discovery engine rather than on the sidelines. Teams work with protein language models (PLMs), 3D structure prediction and architectures like STAG-LLM to understand how proteins and candidate drugs interact. The goal is brutally concrete: shorten the path from idea in a lab notebook to a therapy for epilepsy, Parkinson’s or autoimmune disease, as outlined in UCB’s own overview of AI-enabled drug discovery.

How AI meets biology

The stack is largely Python plus TensorFlow and specialist bioinformatics libraries, but your closest collaborators are often biologists rather than software engineers. Day to day, you might:

• Train PLMs on proprietary sequence and assay data
• Use ML to prioritise compounds for wet-lab experiments
• Build models that support pharmacometrics and clinical trial design

Many roles sit across sites in Brussels, Leuven and Braine-l’Alleud, with cross-border teams in the UK and Germany for pharmacometrics and neurology.

Why it matters in Belgium

Belgium has quietly become an AI-biotech corridor, and UCB is one of the companies that reports on “AI factories” in pharma frequently point to when arguing that the country can lead in data-driven life sciences. External analyses of AI in drug discovery note that approaches like those UCB pursues are starting to fix “some of drug discovery’s most persistent problems”, from high attrition to slow timelines (Drug Target Review’s coverage of AI in drug discovery).

For Belgian ML engineers with a soft spot for biology, UCB offers something few pure software companies can: models that don’t just optimise ads or credit limits, but may one day sit behind a treatment prescribed in a hospital in Brussels, Liège or Ghent.

KBC

Day-to-day impact

Open the KBC app on a train between Leuven and Brussels and you are likely talking to Kate, KBC’s digital assistant. For AI engineers, Kate is the flagship product: a LLM-powered conversational interface that handles everyday banking questions in Dutch, French and English, embedded directly into one of Belgium’s most used banking apps.

Behind that friendly chat bubble sit models for intent detection, dialogue management, sentiment analysis and proactive nudges across retail banking, insurance and investments. Unlike many B2B Belgian AI deployments, you ship features that are touched by tens of thousands of customers every day.

How the stack looks

KBC leans on Azure for cloud infrastructure, with Python as the main implementation language and a growing reliance on large language models for natural dialogue. Engineers work in product squads where they:

• Design and evaluate multilingual prompts and fine-tuned models
• Meet tight latency and availability SLAs for consumer traffic
• Collaborate with risk and compliance on safe, explainable AI behaviour

Interviews are known to probe system design for high-availability apps as much as ML theory, reflecting the scale of KBC’s digital channels.

Why it stands out in Belgium

KBC is one of the few Belgian employers where you can work on production conversational AI at true consumer scale. AI and data roles at KBC’s Brussels and Leuven sites feature prominently in overviews of where companies are hiring AI engineers in Belgium, alongside other large banks and telcos (market guides to AI jobs in Belgium).

Compensation for senior profiles is competitive with other major Belgian banks and insurers, usually including the familiar package of company car, bonus and extensive insurance. If your ideal “praline” is seeing conversational models in the hands of ordinary Flemish and francophone users every single day, KBC’s Kate platform is a strong bet.

Google

What the roles look like

Google’s Belgian presence is small but strategically placed. When an AI-focused role opens - for example, a “Software Engineer, PhD, Early Career, AI/Machine Learning” position - it typically centres on large-scale ML systems using JAX, TensorFlow, Python and C++, and on rethinking how people search for and interact with information globally. The official role outlines stress work on distributed training and inference across Google’s worldwide infrastructure, not just local prototypes in Brussels or Leuven (Google’s AI/ML early-career job description).

Day-to-day engineering

From Belgium, you contribute to products used on every continent. Typical responsibilities include:

• Designing and scaling ML pipelines across distributed clusters
• Optimising model performance under tight latency and reliability constraints
• Collaborating with teams in other European and global offices on new user experiences

The focus is less on domain-specific business logic and more on foundation-level infrastructure that other teams and products will build on.

Getting in the door

The hiring process follows the standard Google pattern: multiple coding interviews, at least one ML theory or system design round, and behavioural conversations probing the famous “Googliness” traits. Strong fundamentals in algorithms and data structures are non-negotiable; deep familiarity with modern ML stacks is expected rather than considered a bonus.

Why Belgium changes the angle

Sitting in Brussels or Leuven, you get a front-row seat on how a hyperscaler adapts its AI to EU regulation, privacy rules and multilingual markets, while enjoying a more manageable cost of living than in places like Zurich or London. Studies on how to “make AI work for Europe” emphasise trust and accountability; working for Google here means grappling with those constraints directly, not as an afterthought.

Microsoft

Where Microsoft Belgium uses AI

In Brussels, Microsoft sits at the junction of engineering, research and government affairs. Local AI-related roles typically orbit around Azure AI Services, enterprise-scale GenAI solutions and co-innovation projects with public administrations and EU-level institutions. Some positions plug directly into international research efforts, as seen in the AI and machine learning roles listed among Microsoft Research open positions.

What AI engineers actually do

Rather than building consumer apps from scratch, you are helping Belgian and Benelux organisations become “AI-native” on top of Azure. Day to day, that typically means:

• Designing reference architectures, prototypes and MLOps patterns using Azure OpenAI, Cognitive Services and Azure Machine Learning
• Embedding Copilot-style experiences into productivity and line-of-business tools
• Implementing guardrails for GenAI safety, compliance and monitoring across multilingual (NL/FR/EN) environments

Many engineers work in hybrid roles that mix hands-on coding with solution architecture and technical advisory work for customers.

Policy, public sector and the Belgian angle

Because so many EU institutions and federal bodies sit within a few tram stops, Microsoft Belgium plays a visible role in debates on AI regulation, data residency and digital inclusion. It is an active partner in initiatives like the national DigitAll ecosystem, working alongside telecoms and banks to close digital gaps and promote responsible AI use in public services.

Why it might belong in your “box of ten”

If you enjoy translating cutting-edge models into stable patterns that other teams can reuse, and you want firsthand exposure to how AI policy is shaped and implemented in Europe, Microsoft Brussels offers a rare combination: deep Azure AI access, direct contact with large Belgian clients and a front-row seat on the regulatory conversations happening just across the canal.

Showpad

Down by the Ghent canals, Showpad is what many engineers picture when they say they want a “real” scale-up: international customers, fast feedback loops, and enough structure that your models don’t die in a notebook. The company builds a sales enablement platform used by B2B teams worldwide, with AI quietly powering which content a rep sees, how meetings are analysed and which deals look wobbly.

What you actually build

AI engineers focus on three main areas: recommendation engines that select the next best piece of sales content, NLP over emails and meetings to gauge sentiment and themes, and predictive models that flag deals at risk so managers can intervene. Showpad leans on a modern stack - Python, PyTorch and cloud data warehousing such as Snowflake - similar to what you’ll find across the more advanced Belgian AI startups highlighted in reports on Belgian AI start-ups and scale-ups.

• Ranking and recommendation models for content and coaching
• Sequence and text models for sales conversations
• Scoring models for pipeline health and forecast accuracy

How engineering feels from the inside

The culture is product-led: ML engineers sit in cross-functional squads with frontend, backend and product, and often own models end-to-end from data pipeline to monitoring dashboards. You see feature usage and impact quickly because sales teams are vocal internal customers; if a recommendation model misbehaves, your inbox will know about it before any dashboard does.

Why it stands out in Ghent

Showpad is tightly woven into the Ghent tech ecosystem and talent coming out of Ghent University, part of the broader wave of AI-driven scale-ups that analysts describe as “the Belgium advantage” for startup innovation (analyses of Belgian AI startups and recruiting). If your ideal praline is somewhere between corporate stability and three-person startup chaos, Showpad offers a balanced, hands-on environment to ship ML into a global SaaS product from a base in Flanders.

Solvay

Take AI out of the browser and into a chemical plant and you land somewhere close to Solvay. As a global chemicals and advanced materials group headquartered in Brussels, Solvay uses machine learning to squeeze more performance from both molecules and machinery, turning fairly abstract models into concrete gains in yield, safety and emissions.

What you actually build

AI engineers here work on three big fronts: predictive maintenance for rotating equipment and critical assets, molecule and materials design for more sustainable products, and end-to-end process optimisation to cut energy use and waste. The stack is dominated by Python, scikit-learn and specialised cheminformatics and process-modelling tools tuned for time-series and spectral data rather than web logs.

• Forecasting failures in compressors, pumps and reactors before they shut a line down
• Building surrogate models of complex chemical processes to explore new operating windows
• Learning structure-property relationships to propose greener formulations faster

How engineering feels on the ground

You split time between Brussels HQ and industrial sites such as Aubange, working with process engineers, operators and materials scientists. Success is measured in fewer unplanned shutdowns, tonnes of CO₂ avoided and new products that make it from pilot line to customer, not just in dashboards.

Why it matters in Belgium

Analyses of Belgium’s AI leadership in industry often highlight Solvay and similar players as anchors of “AI for the physical world”, complementing the country’s strength in pharma and biotech by pushing ML into factories and materials labs (sector reports on AI in Belgian industry).

For senior roles, total compensation typically falls around €80,000-€100,000 with the usual Belgian extras (company car, insurance, bonus). If your ideal praline is about climate impact and industrial problem-solving rather than pure software, Solvay offers a rare chance to apply AI to real reactors and real sustainability targets from a base in Belgium.

How to choose your Belgian AI employer

Choosing your “box of ten” is less about finding the mythical #1 employer and more about deciding which fillings you actually enjoy: research papers or product launches, biopharma or banking, Leuven labs or Brussels high-rises. The same company can be a dream fit for a Ghent-based deep-tech fan and a poor match for someone who really wants consumer apps in French and Dutch.

Start with your own priorities

Before you chase brand names, be explicit about what you want to optimise for over the next 3-5 years: learning curve, research depth, stability, product impact, language mix. Market analysts note that Belgian employers increasingly seek engineers who can stabilise and optimise existing applications, not just build greenfield prototypes, which shifts the balance towards strong MLOps and integration skills (Belgium tech hiring outlook from Source Group International).

Match priorities to hubs and archetypes

Use this list less as a ranking and more as a map. Different Belgian hubs - Brussels, Leuven, Ghent, Antwerp - specialise in different “tastes” of AI work, and the companies in this article roughly cluster into a few archetypes:

Taste before you commit

Where possible, “taste” a company before you buy the whole box: attend meetups they sponsor, contribute to open-source they maintain, or join as an intern or working student. Platforms that track AI companies and startups in Belgium are useful for scouting lesser-known employers that fit your criteria. If you realise a gap - in Python, MLOps or domain knowledge - consider structured upskilling through Belgian universities or affordable bootcamps like Nucamp before making your next move.

Frequently Asked Questions