Research Internship

Publication date : Nov 13, 2026, 8:56PM Contract duration : 6 months What does the position involve? Telecom operators manage data distributed across numerous tables linked by keys, including customer, contract, contact, network incident and service usage data. This relational organization makes it possible to represent customers’ detailed histories, but it makes this kind of data more difficult to use with conventional machine-learning methods. Such methods generally require multi-table data to be transformed into a single table through handcrafted aggregation operations (Mean, Min, Max…) and feature engineering. The objective of this internship is to study and compare two classification strategies for this type of data. The first will rely on an established automatic propositionalization approach, notably using the Khiops tool [1], which automatically builds aggregated variables from secondary tables. The second will rely on recent relational foundation models, such as KumoRFM-2 [3], which represent a relational database as a heterogeneous and temporal graph: table rows are represented as nodes, while links between primary and foreign keys form the edges. To ensure reproducible evaluation and position the results with respect to the state of the art, the internship will include experiments on the public RelBench benchmark [2]. This benchmark provides relational databases, predictive tasks and standardized evaluation protocols for machine-learning models applied to this type of data. Synthetic data may also be considered for specific experiments. “In-context” foundation models such as KumoRFM-2 appear promising on small databases, but their ability to scale to large databases remains to be demonstrated. Attention will therefore be paid to identifying situations in which a frugal and explainable multi-table method such as Khiops is sufficient, as well as situations in which directly learning the relational structure with a foundation model such as KumoRFM-2 could make it possible to exploit more subtle temporal or multi-table interactions. References [1] Boullé, M., Charnay, C., & Lachiche, N. (2019). A scalable robust and automatic propositionalization approach for Bayesian classification of large mixed numerical and categorical data. Machine Learning, 108(2), 229–266. [2] Robinson, J., Ranjan, R., Hu, W., Huang, K., Han, J., Dobles, A., Fey, M., Lenssen, J. E., Yuan, Y., Zhang, Z., He, X., & Leskovec, J. (2024). RelBench: A Benchmark for Deep Learning on Relational Databases. Advances in Neural Information Processing Systems. https://cs.stanford.edu/jure/pubs/relbench-neurips24.pdf [3] Hudovernik, V., López, F., Kocijan, V., Nitta, A., Lenssen, J. E., Leskovec, J., & Fey, M. (2026). KumoRFM-2: Scaling Foundation Models for Relational Learning. arXiv preprint arXiv:2604.12596. Who are we looking for? You are currently completing a Master’s degree or an engineering degree, with a specialization in deep learning and artificial intelligence. You have a good command of Python and Scikit-learn. You have a solid background in statistics and mathematics, as well as an interest in both the applied and theoretical aspects of the subject. You have a B2 level of English, both written and spoken. You are curious, open-minded and autonomous. You enjoy working as part of a team and are willing to share your ideas, results and questions. Ideally, you are considering pursuing a PhD in a related field. The possibility of continuing with a PhD within the team may be considered. This is a six-month internship starting in January or March 2027. Please note that we do not offer work-study internships or internships involving three or more interruptions. Internships cannot be converted into work-study contracts; the work-study recruitment campaign will begin around March. Why join Orange?
• You will join a high-level research and development team whose expertise in statistical data processing and machine learning is recognized within Orange and internationally;
• You will work on concrete operational challenges and real-world data;
• You will collaborate with Orange France’s operational units;
• Your work may lead to a scientific publication;
• The possibility of pursuing a PhD within the team may be considered;
• You will benefit from Lannion’s exceptional quality of life. The proposed gross monthly internship allowance ranges is €1,680 for a Master’s-level internship. Additional benefits include:
• Access to the social and cultural activities offered by the Social and Economic Committee;
• Access to the company restaurant or meal vouchers;
• Reimbursement of 50% of the cost of a public transport pass. About the working environment The PROF team conducts research in statistics, machine learning and data analysis. It provides its expertise in these fields to support Orange’s operational units. Desired start date : Mar 01, 2027, 12:00AM At Orange, only your skills matter. Regardless of your age, gender, background, origin, religion, sexual orientation, disability, neurodiversity, or appearance, we actively encourage diversity within our teams, as it is a collective strength and a driver of innovation.
Orange is a disability-inclusive employer: please feel free to let us know about any specific needs you may have.