Unlock Premium Benefits

Get Immediate Access to 3,991,000 premium jobs

- Direct Access to Top Employers

Gain full access to exclusive job listings from leading companies worldwide.

- Verified, High-Quality Jobs Only

No ads, scams, or junk-just genuine opportunities.

- Focus on Real Opportunities

Explore thousands of open positions tailored to your lifestyle, including adaptable remote jobs.

- Exclusive Resume Review

Receive expert feedback with personalized suggestions to enhance your resume.

Continue →

Description

The general aim of the project is to investigate and do research around the role of mutual information and certain information-theoretic operators in self-supervised learning (SSL), particularly in the pretraining phase of certain large language models (LLMs). Building on state-of-the-art research, and the (new) probabilistic view of probabilistic inference, the project seeks to develop both theoretical and empirical insights into the relationship between mutual information maximization, predictive probabilistic models, and the SSL performance.

Line of research:

Machine Learning; Artificial intelligence; Probabilistic models and Bayesian inference; Self-supervised learning and information-theoretic operators.

Tasks to be performed:

The successful candidate will contribute to research on the role of mutual information in self-supervised learning (SSL), with particular emphasis on large language model pre-training. The work will focus on the theoretical understanding of SSL objectives and their connections to mutual information maximization, probabilistic modelling, Bayesian inference, and uncertainty quantification.

The main activities will include:

- Reviewing the literature and state of the art on information-theory, masked SSL, probabilistic modelling, and Bayesian inference.

- Investigating theoretical connections between SSL objectives, information-theoretic operators,



conditional prediction and probabilistic models.

- Developing novel probabilistic and Bayesian frameworks to better understand SSL performance and quantify uncertainty.

- Designing and conducting computational experiments (i.e. for natural language processing (NLP) or computer vision (CV)) to validate theoretical findings and evaluate uncertainty quantification methods during pretraining.

- Contributing to scientific publications, research meetings, and presentations of results.

The duration of these tasks is estimated over a period of 6 months.

Group and complement: Group 3 + Level u

Dedication and working hours: Full time (35h/week)

Planned remuneration approx: 29,979.98 € gross per year

Financing fund: Supported by Junior Leader Fellowship Programme 2024 "La Caixa" Banking Foundation. (Grant Agreement ID: LCF/BQ/PI24/12040025)

Qualifications

Requirements:

- Batxillerat-FP2/ High school-FP2

The expected start date is from December 1, 2026, the job is in Barcelona.

Selection criteria:

The selection of the candidates will be made through evaluation of the curriculum and, where appropriate, with the carrying out a test and/or interview. Valuation will be as follows:

1- Academic Training (0-40 points).

- Bachelor and / or Master's degree in Artificial Intelligence or areas related to ML/AI (i.e. computer science, telecommunication engineering, or data science). In case the candidate has not completed the Bachelor degree and/or master's studies, it will be assessed whether he is doing the final bachelor, master's thesis or similar,



and its subject.

- Having taken courses or participated in projects related to machine learning, computer vision and natural language processing (NLP) during the bachelor and master’s degrees.

2- Other professional training and experience, adequacy to the proposed profile (0-45 points)

- Knowledge in machine learning methods in general, and probabilistic methods, uncertainty quantification and Bayesian inference in particular.

- Knowledge or experience with information theory would be a plus.

- High level of programming skills in Python and using of large-scale computing tools (HPC, GPUs, etc).

- Experience in using simulation/experimentation for research in the areas of computer vision and NLP.

- Experience in empirical evaluation techniques, statistical analysis, reproducible simulations and computer-based experimentation in general.

3- Other merits (0-15 points)

- Advanced level of English (spoken, written). Demonstrated proficiency.

- Outstanding interest for AI/ML problems.

- Experience in research and/or writing scientific documents.

The minimum score to pass the selection process is 80 points. The candidate with the highest score in the selection process will be offered the job.

Application Instructions

BASIC INFORMATION ON DATA PROTECTION:

Data controller: Universitat Pomeu Fabra

Purpose: Research support staff recruitment.

Rights: You can access your data; request their rectification, deletion and portability; you may object to their processing and request their limitation.

Additional information: Further detailed information is available on the website: https://rat.upf.edu/ca/ll040

Equal Employment Opportunity Statement

UPF promotes a diverse and inclusive environment and welcomes applicants regardless of age, disability, gender, nationality, race, religion or sexual orientation.

📌 ENG-PSR--39 - Specialist support technician: Probabilistic prediction and information-theoretic operators for SSL (Barcelona)
🏢 Universitat Pompeu Fabra
📍 Barcelona