VDAB’s mission is to help every job seeker find their dream job.

We helped improve the job seeker experience at the Dutch public employment service, VDAB. By offering personalized recommendations, we helped increase the number of job applications through their various digital channels.



Our client’s mission is to help every job seeker find their dream job. Our solution is helping them achieve that.


A job recommender has to predict how likely a job seeker will be interested in a vacancy, based on everything it knows about that job seeker and that vacancy. We built a Deep Learning model that accurately predicts job interest, by learning from millions of historical interest signals.

Classical, rule-based systems achieve this goal by manually defining matching criteria. For example: “To work as an engineer, you need an engineering degree”. This is a fragile system, which often doesn’t return the most relevant suggestions. What if, instead of defining the rules manually, we could learn those rules from the millions of historical examples of job seekers who applied for certain jobs? This is exactly what we did. We built a Deep Learning model that takes as input profiles of job seekers, job vacancies, and historical interest signals between them. The model learns to predict these interest signals. For example: “I often see Psychology majors applying for HR-related jobs, so I will remember this relation because this helps me correctly predicting interest signals”.

Success criteria

Improved KPI’s versus the current systems. We want the job suggestions to be of higher relevance, resulting in more clicks and eventually more applications. Apart from that, the system needs to be cross-lingual and retrained automatically.


Superb performance on the hold-out data set. Using a model trained on vacancy clicks, we can accurately predict vacancy clicks of today for any job seeker. Our job recommender helps thousands of job seekers find better jobs on a daily basis. We’re continuously monitoring and improving the KPIs