Job offer

• Build and openly publish a collection of GNNs trained on tasks with known answers (minimum paths, flows, spanning trees), to validate interpretability methods.
• Adapt to GNNs the interpretability tools created for large language models, such as sparse autoencoders, to locate the variables that the network represents internally in nodes and edges.
• Design causal intervention experiments (e.g. activation patching) to check which of these variables the network actually uses, and choose the most compact description that remains faithful to its behavior.
• Apply the methods to real GNNs that act as digital twins of communications networks (such as RouteNet) and other sectors (energy, traffic, agri-food chain).
• Publish the results in top-level conferences and journals on machine learning and networks, release the code in open access and actively participate in the international interpretability community., * The maximum score for the candidates who accomplish all the eligibility criteria stated in the job offer will be 10 points.
• Threshold: 5 points.
• The maximum score (10 points) will be distributed as follows:
• 1 point for required speciality.
• 2 points for required academic training.
• 2 points for technic competences.
• 1 point for organisational competences.
• 3 points for professional experience.
• 1 point for any aspect of the candidate’s professional profile to be determined as specially relevant by the Selection committee.

If the selection committee decide to include a personal interview as additional step of the selection process, only the candidates scored 5 or above for their CV will be retained for the interview step.

The interview will be assessed according the following criteria:

• Maximum scoring: 5 points.
• Threshold: 3 points.
• The maximum score will be distributed according the following criteria:
• 2 points for adequacy to functional competences of the job.
• 2 points for adequacy of professional experience.
• 1 point for any aspect of the candidate’s professional profile to be determined as especially relevant by the Selection committee.

The eligible candidates will be ranked from highest to lowest score, being this the selection criteria. Selection process

The planned selection process is as follows:

• Those interested must access the website where the university publishes all the offers: https://talenthub.upc.edu/en/jobs/r1/r1-jobs/r1-jobs, select the offer that is of interest to you, click on: https://seuelectronica.upc.edu/en/procedures/call-for-recruitment-of-research-staff-in-training-predoctoral?set_language=en, fill in the form and add the additional documentation indicated.
• The members of the selection committee will decide for each case what the selection process will be, if an assessment is made only for the curriculum vitae or any other procedure is planned: interview, tests, etc …
• Each phase of the selection process will be reported on the website where the offer is published. If applicable, indicate your request.
• Finally, the name of the person selected will be published on the same website where the offer is published.

Master Degree or equivalent

Research Field Engineering » Communication engineering

Education Level Master Degree or equivalent

Research Field Mathematics » Applied mathematics

Education Level Master Degree or equivalent

Research Field Physics » Computational physics

Education Level Master Degree or equivalent

Skills/Qualifications

Technical Skills:

• Advanced programming in Python, PyTorch, PyTorch Geometric or DGL.
• Version control (git), Linux and model training on GPU (reproducible experiments).

Other requirements to consider:

• Curiosity to understand how AI models work internally, ability to work independently and willingness to work in a team.
• Availability to undertake international research stays and to follow specialized training programs (e.g. ARENA).

Specific Requirements

Knowledge:

• Linear algebra, probability and statistics.
• Graph theory and algorithms on graphs.
• Machine learning and deep learning (neural networks, model training and evaluation).
• Graph neural networks (GNN) or other current architectures (e.g. transformers).
• Causal inference or computer networks.
• Interpretability of AI (mechanistic or explainable).

Professional Experience:

• Analysis of scientific literature.
• Contributions to open source projects, machine learning competitions or relevant own projects.
• Have participated/collaborated in research projects.
• Experience in functions similar to those described will be valued, specifically in the development of research activities, both in the university and industrial environment.

Annual gross salary:

• 1st year: €20,364.96
• 2nd year: €20,364.96
• 3rd year: €25,456.20

Eligibility criteria

All the candidates who cannot prove the required academic degree will be immediately withdrawn from the selection process.