Matching Similar Employee Competences with AI

Client's challenge

• VDAB has a growing database of over 11,000 competencies, skills, and knowledge elements.
• Avoiding duplicate competencies and grouping similar ones manually is no longer feasible due to the database's size.
• They needed an automated solution to link similar competencies effectively.

Our Solution

• Developed a specific algorithm utilizing the Google Translate API to translate competencies from Dutch and French to English, resulting in four different language versions.
• Employed word embeddings and both classical and machine learning-based methods for sentence embeddings.
• Ranked each sentence embedding and retrained models based on continuous feedback from VDAB to ensure accurate matching.
• Created a custom-built user interface to facilitate continuous feedback and algorithm updates.

Business impact

• Enabled VDAD to quickly clean up their database and search for matching competencies.
• The feedback loop in the user interface keeps the algorithms up to date, ensuring ongoing accuracy.
• Significantly reduced the time required for VDAB to perform search queries, overall enhancing efficiency.