Missions

The candidate engineer will have three objectives:
Generate training databases for upskilling using alignment methods from the literature in Reinforcement Learning from Human Feedback: Direct Preference Optimization (DPO) [24] and Kahneman-Tversky human utility model (KTO) [34], Preference Policy Optimization algorithms (PPO) [26], preference learning (ORPO) [16], Alignment via Optimal Transport (AOT) [22].
Implement a fine-tuning pipeline that makes use of those databases to reproduce pedagogical strategies as LMs. Pedagogical strategies include scaffolding, mastery learning and collaborative learning.
Work with a post-doctoral researcher to deploy controlled experiments and collect well-being dimensions.

Activités

Implement and deploy a fine-tuning pipeline
Compare the effectiveness of those pipelines for upskilling
Design and implement plugins for two educational platforms
Deploy controlled experiments designed by a postdoctoral researcher to gather well-being dimensions
Take active part in the project deliverables and weekly technical meetings

Compétences

Abstraction capabilities, strong programming skills in C/C++ and Python, familiarity with frontend design, collaboration skills. English is needed.
Hybrid skillset, combining advanced programming capabilities (e.g., experimental design, innovation) with strong software engineering expertise (e.g., scalability, production deployment, optimization).

Contexte de travail

The performance of recommendation algorithms that make use of human behavior heavily depends on their ability to capture the experience of people who interact with them. For instance, in Education, recommendations of courses and tests should consider not only the learners' evolving knowledge but also aspects of their well-being such as engagement, motivation, interest, satisfaction, frustration, boredom, fatigue and anxiety [0]. Various studies have shown that integrating well-being into decision-making generates better mental health, less burnout, and better performance over time [3,6]. The work of the candidate engineer will be in the context of FeelGoodAI, a project with the goal of rethinking recommendation approaches to represent and leverage human well-being in decision-making.

The work will take place at the Grenoble Informatics Lab (LIG), a 450-member laboratory with teaching faculty, full-time researchers, PhD students, administrative and technical staff. The mission of LIG is to contribute to the development of fundamental aspects of Computer Science (models, languages, methodologies, algorithms) and address conceptual, technological, and societal challenges. The 22 research teams in LIG aim to increase diversity and dynamism of data, services, interaction devices, and use cases influence the evolution of software and systems to guarantee the essential properties such as reliability, performance, autonomy, and adaptability. Research within LIG is organized into 5 focus areas: Intelligent Systems for Bridging Data, Knowledge and Humans, Software and Information System Engineering, Formal Methods, Models, and Languages, Interactive and Cognitive Systems, Distributed Systems, Parallel Computing, and Networks.

The host team, DAISY, is a joint CNRS, Grenoble INP, and UGA research team handling research challenges at the intersection of AI and data management, but also when data is sourced from interdisciplinary domains such as education and health.
The position is located in an area subject to French legislation on the protection of scientific and technical potential (PPST), and therefore requires, in accordance with regulations, that your arrival be authorized by the competent authority of the Ministry of Higher Education and Research (MESR).

Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.

Contraintes et risques

N/A