Missions

We are seeking two Machine Learning Engineers to join our team. You will be
responsible for the design, implementation, and empirical evaluation of
metacognitive and self-reflective mechanisms in AI architectures. Your work will
support our research program to develop and test models of self-monitoring,
uncertainty estimation, and introspection in artificial systems.

Activities

- Test and fine-tune pretrained LLMs.
- Design new ML models incorporating metacognitive and self-reflective
components.
- Design benchmark and testing protocols for introspective performance in AI
agents.
- Analyze results and contribute to publications and conferences.
- Maintain clean, well-documented, and reproducible code.

Skills

- Master's or PhD in Computer Science, Machine Learning, AI, Computational
Neuroscience, or related field.
- Strong experience with large language models (LLMs), including fine-tuning,
and applying them in real-world tasks.
- Solid experience in deep learning with the PyTorch framework.
- Excellent coding skills with a strong ability to prototype and iterate quickly.
- Ability to work in a multidisciplinary team while maintaining a high level of
autonomy.
- Familiarity with uncertainty estimation, Bayesian methods, self-supervised
learning, and/or meta-learning (plus but not required).
- Interest in theoretical and philosophical questions around consciousness and
metacognition (plus but not required).

Work Context

The Brain & Consciousness team at École Normale Supérieure investigates the
neural and computational foundations of consciousness, metacognition, and
self-reflective méchanisms. Our interdisciplinary approach combines
neurotechnologies, cognitive neuroscience, philosophy of mind and cutting-edge
machine learning. This project is a central pillar of the team's new effort in designing
AI systems inspired by human self-awareness, aiming to integrate these insights into
real-world machine learning applications. The team is led by Pr. Sid Kouider, CNRS
research director and former CEO of NextMind (acquired by Snap).