Carlos Stein
Naves de Brito.

Director · NightCity Labs · Lisbon

Independent AI-driven research and discovery lab developing multi-agent frameworks, with applications across computational neuroscience, machine learning, reliable AI, AI process automation, robotics, consciousness, and NeuroAI.

Research Scientist · Champalimaud Research

Natural Intelligence Lab, Champalimaud Centre for the Unknown (CCU), Lisbon, Portugal.

Visiting Scientist · EPFL

Laboratory of Computational Neuroscience (LCN), Lausanne, Switzerland.

Postdoctoral Researcher · Gatsby UCL

Gatsby Computational Neuroscience Unit, University College London. Supervisor: Peter Latham.

Machine Learning Expert · Cambridge Spark · Cambridge, UK

Lead developer of an advanced machine learning course for hedge-fund scientists at G-Research, covering ML fundamentals, deep learning, large-scale recurrent models, and Bayesian ML. Lead developer of the public G-Research Crypto Forecasting Kaggle competition (a Featured competition with 23,000+ entrants, ~1,900 teams). Authored a widely-used tutorial notebook on robust time-series pipelines and validation schemes.

Software Engineer Intern · Google · Mountain View, CA

Software development for large-scale data storage and retrieval in online advertisement systems.

Software Engineer Intern · Aero-Spatial Institute (IAE), CTA · Brazil

Development, parallelisation and optimisation of high-performance software for computational fluid dynamics.

PhD · Computational Neuroscience · EPFL

Laboratory of Computational Neuroscience, École Polytechnique Fédérale de Lausanne, Switzerland. Supervisor: Wulfram Gerstner. Dissertation: Theory of representation learning in cortical neural networks.

MSc · Neuroscience · University of São Paulo

Laboratório de Neurociências – Hospital Sírio-Libanês, USP, Brazil. Supervisor: Koichi Sameshima. Dissertation: Statistical models of connectivity between brain areas.

BSc · Computer Engineering · Instituto Tecnológico de Aeronáutica (ITA)

Brazil.

Brito C.S.N.. arXiv:2603.01184 [cs.LG].

Brito C.S.N.. arXiv:2602.11090 [cs.LG].

Brito C.S.N. et al.. ICLR 2026, Post-AGI Science and Society Workshop.

Brito C.S.N.. arXiv:2506.19726 [cs.LG].

Brito C.S.N.. arXiv:2508.15019 [cs.LG].

Brito C.S.N.. International Conference on Machine Learning (ICML).

Brito C.S.N. et al.. Advances in Neural Information Processing Systems (NeurIPS).

World Models as Reference Trajectories for Rapid Motor Adaptation

Brito C.S.N. et al.. ICLR Robot Learning Workshop.

Brito C.S.N., Gerstner W.. PLOS Computational Biology 20(2): e1011844.

Brito C.S.N., Gerstner W.. arXiv:2105.10109.

Brito C.S.N., Gerstner W.. PLOS Computational Biology 12: e1005070.

Unified Asymptotic Theory for All Partial Directed Coherence Forms

Baccalá L.A., Brito C.S.N., Takahashi D.Y., Sameshima K.. Philosophical Transactions of the Royal Society A 371: 20120158.

Asymptotic behavior of generalized partial directed coherence

Brito C.S.N., Baccalá L.A., Takahashi D.Y., Sameshima K.. Annual International Conference of the IEEE EMBS, pp. 1718–1721.

The Blue is Sky: Color Qualia as Learned Associative Structures

Association for the Scientific Study of Consciousness (ASSC).

Hierarchical Integration of RL and Cerebellar Control for Robust Flexible Locomotion

Conference on Reinforcement Learning and Decision Making (RLDM) — selected talk.

Dual-Model Framework for Cerebellar Function: Integrating Reinforcement Learning and Adaptive Control

Cosyne.

Plasticity of adaptation regulates neural activity and temporal correlation for arbitrary input statistics

Brito C.S.N., Gerstner W.. AREADNE.

Plasticity of adaptation stabilizes neural activity and induces temporal decorrelation

Brito C.S.N., Gerstner W.. Cosyne.

Homeostatic synaptic depression explains receptive field development by implicit whitening

Brito C.S.N., Gerstner W.. Cosyne.

Nonlinear Hebbian learning as a universal principle in unsupervised feature learning

Brito C.S.N., Gerstner W.. ICML Deep Learning Workshop — selected talk.

A unifying theory of receptive field development

Brito C.S.N., Gerstner W.. Cosyne — selected talk.

A model of perceptual discrimination under sequential sensory evidence

Hughes E., Brito C.S.N., Gerstner W., Romo R., Deco G.. CNS.

General conditions for spiking neurons and plasticity rules to perform independent component analysis

Brito C.S.N., Gerstner W.. CNS.

Generalization, regularization and Bayesian neural networks

Gatsby UCL Transfer Talk.

Adaptive regularization by noise in deep neural networks

Albert Einstein Hospital – Big Data Initiative.

Five ways to regularize deep neural networks

LCN EPFL Seminars.

An empiricist connectionist theory of color perception

Herzog Lab – EPFL Seminars.

Theory of synaptic plasticity for receptive field development

Oxford Neurotheory Seminars.

Sparse coding with spiking neurons and Hebbian plasticity

FACETS-ITN meeting.

Camillo Heye · MSc thesis project

CCU / LMU Munich. Modelling naturalistic motor control and its dysfunction with deep reinforcement learning.

Eszter Vertes · MSc thesis project · EPFL

Top-down mechanisms in the visual cortex. Best thesis award.

Gevorg Poghosyan · Master semester project · EPFL

Multilayer unsupervised learning with realistic networks.

Guillaume Riesen · Fulbright project · EPFL

Development of V1 orientation maps with local plasticity rules.

Graduate Course Teaching Assistant · EPFL

Unsupervised and reinforcement learning in neural networks · pattern classification and machine learning · biological modeling of neural networks.

Tutor · Cajal Course in Computational Neuroscience (CCCN)

Champalimaud Centre for the Unknown, Lisbon.

Tutor · No-Budget Science Hack Week

Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro.

Silver Medal · International Mathematical Olympiad (IMO)

Washington DC, representing Brazil.

World Finalist · International Collegiate Programming Contest (ICPC)

Gold Medal ×4 · Brazilian Mathematical Olympiad

2000, 2001, 2003, 2004.

Second Prize · International Mathematics Competition (IMC)

University-level.

First Prize · Brazilian Physics Olympiad

Brains, Minds & Machines (CBMM) Summer Course

Woods Hole, USA.

Advanced Course in Computational Neuroscience (ACCN)

Będlewo, Poland.

FACETS-ITN · Marie Curie PhD scholarship

DAAD Hochschulwinterkurs scholarship for German studies

Freiburg, Germany.

Conference Media Chair · Cosyne

Created the Cosyne Talks YouTube channel, extending the conference to a global audience with live streams during each meeting and an open archive of past talks. cosyne.org.

External Seminar Organiser · Champalimaud Research

Founder and Organiser · Theory Mini-Symposium series · Champalimaud Research

Seminar Organiser · Gatsby UCL Weekly External Seminars

Referee · Cosyne · NeurIPS · ICML · ICLR

Guest editor · PLOS Computational Biology

Chatsubo · Interactive AI art installation

Metamersion: Healing Algorithms, Champalimaud Foundation, Lisbon. Immersive multi-agent conversational environment exploring identity and human–AI interaction.

Science on the Walls · Ciência nas Paredes

Volunteer scientist, hands-on science and art activities for children in underprivileged neighbourhoods of Lisbon.

Artificial Intelligence and Medicine · seminar

Itaú Cultural, São Paulo, Brazil.

Neuroscience: the search for understanding brain and mind

Book chapter with O. Pessoa, in Science at the turn of the century, EDUFBa, Brazil.

How the brain represents the world · seminar

Arte della Tavola art exposition, Lausanne, Switzerland.

Portuguese (native) · English (fluent) · Spanish (advanced) · French (intermediate) · German (intermediate)

Python · C++ · Julia · Matlab · PyTorch · Google ADK (multi-agent RL framework)