The AI recruiting assistant RABE: An R&D project with SRH University Heidelberg
As part of a collaboration between ONTEC AI and SRH University Heidelberg, students work on solutions for real AI problems. The goal is to directly transfer the valuable ideas of students into practice. One of the projects is the recruiting assistant RABE. This AI-based tool analyzes data from job boards and helps non-technical recruiters identify necessary skills for technical jobs.
About SRH University Heidelberg
SRH University Heidelberg is a private university known for its practice-oriented teaching and modern research approaches. With degree programs in areas such as data science and artificial intelligence, it offers a solid education for future professionals.
The university emphasizes the connection between theory and practice, providing students with practical experiences. Its international orientation and focus on applied research make it a suitable partner for companies in the AI software industry.
SRH University Heidelberg and ONTEC AI decided to collaborate to combine academic knowledge and industrial practice. This cooperation allows ONTEC AI to benefit from the university’s scientific expertise, while students gain practical experience in developing AI solutions.
The challenge: the variety of technical competencies
Recruiting technical talent is often not easy: HR professionals and recruiters often do not know all the technical terms and concepts needed to describe and evaluate job profiles and resumes. This leads to problems in selecting suitable candidates.
The project aimed to help recruiters identify the necessary skills for technical positions and ultimately search for and select them better.
The solution: the recruiting assistant RABE
Based on a briefing from ONTEC AI, the student group had several months to come up with an AI-based solution for the problem at hand.
This resulted in the recruitment assisting buddy engine, also known as RABE. (By the way, the word “Rabe” is German for “crow”)
RABE automates the analysis of job descriptions from online job boards. It shows which skills are needed for specific job titles. Frequently required skills in combination are also recorded.
The tool thus makes it easier for recruiters to match candidates with job requirements and make the best selection.
For this purpose, job descriptions from Austrian job boards are analyzed. RABE uses named entity recognition (NER) to extract relevant skills. Subsequently, a correlation analysis is performed to identify frequently co-occurring skills.
RABE collects data for its analysis from the most common job boards such as metajob.at, karriere.at, and itstellen.at.
Project phases and implementation
The project is divided into two phases:
- In the first phase, the proof of concept, a Jupyter notebook was created. It demonstrates the extraction of relevant skills for specific job titles. Frequently co-occurring skills were also identified. This helped recruiters expand their search criteria.
- In the second phase, a fully functional prototype was developed. This is hosted in a Docker container and has a browser-based user interface. Features include customization of the jobs to be searched, automated nightly updates, and notifications of system issues.
The implementation is done in Python. Libraries such as Numpy, Pandas, Spacy, Tensorflow, FastAPI, and Uvicorn are used. The system uses Docker for containerization, based on the Red Hat ubi8 image.
The system consists of a backend that retrieves and analyzes data, and a frontend. The frontend allows access to the tool. The backend includes NER for skill extraction and a correlation analysis. The frontend is easy to use and provides useful insights.
Results: AI-powered recruiting
RABE increases the efficiency of the recruiting process by reducing the time recruiters spend researching job requirements.
The tool provides a comprehensive list of required and frequently co-occurring skills. This improves the accuracy of candidate searches, allowing recruiters to focus more on evaluating candidates and making targeted selections.
The solution is scalable and can be adapted to additional job boards and job categories.
Follow-up project: personalized job search with AI
Another follow-up project of the collaboration between SRH University Heidelberg and ONTEC AI was the “personalized job search.” This project focused on identifying suitable job offers for candidates from the perspective of recruiters.
The idea was that if an applicant has already gained experience in areas A and B, they might also be suitable for a position in area C.
To analyze these relationships, the project used data from Kaggle, a platform where data scientists work on various challenges. Anonymized resumes were used to analyze candidates’ skills and experiences without relying on real individuals.
This allowed for privacy-compliant yet meaningful analyses to help recruiters identify suitable candidates for open positions.
Summary and conclusion
AI can be used in recruiting in various ways, such as facilitating the selection of particularly promising applicants. An example of this is the RABE project by ONTEC AI, which supports the HR team in analyzing technical skills.
The RABE project not only provides valuable tools for recruitment but also offers students practical experience in solving real problems.
The collaboration between the ONTEC AI team and SRH University Heidelberg demonstrates the value of partnerships between industry and academia.