CV

Education

PhD in Neuroscience, International Neuroscience Doctoral Program, Champalimaud Foundation, Lisbon, Portugal, started September 2025
- Fully funded doctoral program in Neuroscience
MSc in Neuroscience, Faculty of Medicine, University of Lisbon, Portugal, 2022 - expected May 2026
- Elective courses in machine learning and theory of computation
- Curricular average grade of 18 (scale 0-20); defense expected May 2026
Integrated Master’s Degree in Medicine (MD), NOVA Medical School, Lisbon, Portugal, 2013 - 2019
- Final grade of 17 (0-20)
- Year abroad (Erasmus program) at the Charité Universitätsmedizin Berlin, Germany, 2017-2018

March - April 2026: Doctoral Program Rotation Project
- Champalimaud Foundation – Neuropsychiatry Lab, Lisbon, Portugal
- Topic: “Mapping LLM Decision-Making in Computational Psychiatry Parameter Space”. Self-proposed project focused on computational modeling of large language model behavior in a decision-making task, and comparison with healthy and clinical human populations.
- Contributions: experiment design, prompt engineering, implementation of task evaluation pipeline and user interface, computational modeling of behavior.
September 2023 - September 2025: Master’s Thesis Project
- Champalimaud Foundation – Natural Intelligence Lab, Lisbon, Portugal
- Thesis project investigating cognitive alignment between humans and large language models in a collaborative memory foraging task. Co-supervised by Daniel McNamee (Champalimaud Foundation) and Bruno Miranda (University of Lisbon).
- Contributions: design and implementation of online experiments involving human–AI and human–human interaction, data collection (through Prolific), data analysis and manuscript writing. Additional exploration of physiological signals as contextual information for closed-loop human-computer interaction.
- Thesis research produced a NeurIPS 2025 Mechanistic Interpretability workshop poster and a manuscript accepted for publication in PNAS.

2022 - 2023: General Practice, Amadora Health Centre (UCSP Amadora), Amadora, Portugal
- Physician in family medicine clinic (12 months)
2020 - 2021: General Training Internship, West Lisbon Hospital Centre (CHLO), Lisbon, Portugal
- Physician in General Training (12 months)

October 2024 - January 2025: AI Safety Fundamentals, BlueDot Impact (online)
- Online course covering technical AI alignment, including reinforcement learning from human feedback, scalable oversight and mechanistic interpretability.
September 2021: Computational Psychiatry Course, Translational Neuromodeling Unit, University of Zurich and ETH Zurich (online)
July 2021: Neuromatch Academy (interactive track), computational neuroscience course (online)

Lacosse, E.*, Duarte, M.*, Todd, G., Todd, P.M., McNamee, D.C. Artificial Intelligence Models Can Track and Collaboratively Modulate Human Memory Search Dynamics. (in press, 2026) Proceedings of the National Academy of Sciences (PNAS). (*Equal contribution)

Duarte, M., Lacosse, E., Todd, P., McNamee, D.C. Artificial intelligence models can track and collaboratively modulate human memory search dynamics. 7th International Conference on the Mathematics of Neuroscience and AI (neuroMONSTER), June 2026, Rome, Italy. (Upcoming spotlight talk)
Mainen, Z.F., Stein Brito, C., Duarte, M., Sandru, A.R., Neelabh, K., Gingeira, M.L., Zuo, M., Decaix, P., Assaf, D.Y., Xu, A.Z. Thinking About Thinking With Machines That Think. Post-AGI Science and Society Workshop at ICLR 2026, April 2026, Rio de Janeiro, Brazil. (Workshop paper)
Lacosse, E., Duarte, M., Todd, P., McNamee, D.C. Emerging Human-like Strategies for Semantic Memory Foraging in Large Language Models. Mechanistic Interpretability Workshop at NeurIPS 2025, December 2025. (Workshop paper)
Duarte, M., Lacosse, E., Todd, P.M., McNamee, D. Facilitating Cognitive Synergy Between Humans and AI. Responsible AI Forum, November 2024, Porto, Portugal. (Oral presentation, winner of the SPARK award)

Programming: Python (proficient), C#, JavaScript
Machine Learning: Scikit-learn, PyTorch
Natural Language Processing: large language models, embedding models (BERT, FastText)
Mechanistic Interpretability: linear probes, steering vectors, logit lens
Behavioral Studies: experiment design, online data collection (Prolific, Empirica)
Data Analysis: statistical analysis, computational modeling, signal processing
Languages: Portuguese (native), English (C2), German (B2)

November 2024: SPARK award for best MSc research in responsible AI, Center for Responsible AI

May 2024: Chatsubo: an immersive mixed-reality installation
- Metamersion: Healing Algorithms exhibition, Lisbon, Portugal
- Development of a mixed reality installation involving real-time multi-agent interactions between human visitors and large language models, including speech-to-text and text-to-speech functionality and immersive 3D scenes built in Unity.