01 · Current
Three concurrent Geneva appointments connect senior psychiatric care, causal modelling and doctoral research.
Geneva University Hospitals
Senior hospital physician in psychiatry; psychogeriatric crisis programme since 2025. Studied psychedelic-assisted psychotherapy and its associations with Schema Therapy constructs and processes from 2023 to 2025.
Geneva University Hospitals
Causal deep learning for multimodal psychiatric data and real-world psychedelic-assisted-psychotherapy studies. Led a React Native, React, Node.js, Python and R research platform from 2021 to 2022.
University of Geneva Faculty of Medicine
Tests the causal validity of deep sequential task-fMRI models under Professor Stefan Kaiser. The Doctorate in Medicine remains in progress.
02 · Doctorate
The doctorate tests whether forward, backward, autoregressive and masked Transformer models recover planted generative signal or exploit confounding, selection and nuisance information.
Methods Structural causal simulation · task-fMRI · model attribution and interpretability · causal correction · known-truth benchmarking
Babylon Health
Probabilistic inference, deep reinforcement learning, simulated environments, sensitivity and safety methods for triage and active diagnosis.
[RE]MEDs Research
Graph causal methods, simulation-driven validation and target-trial-emulation infrastructure for biomedical research.
Doctorate
Known-truth evaluation of deep sequential models, attribution and causal correction in task-fMRI.
03 · Medical AI
Senior clinical practice, causal multimodal modelling, and psychedelic-assisted-psychotherapy studies of outcomes, safety, biomarkers and implementation.
Sequence-to-sequence models at BIOS Health, with Bayesian optimisation and reinforcement learning for peripheral-nerve stimulation.
Bayesian models, deep reinforcement learning and simulated environments at Babylon Health for triage and active diagnosis; first-author research and three patent families.
This site presents research and training software. It makes no claim of a validated diagnosis, treatment or clinical device.
04 · Experience
[RE]MEDs Research
Led hybrid language-model and biomedical knowledge-graph research for drug repurposing; developed graph causal methods, simulation-driven validation and target-trial-emulation infrastructure; built Python, React and Node.js research tools.
BIOS Health
Developed sequence models for peripheral-nerve and physiological signals, applied Bayesian optimisation and reinforcement learning, and contributed Swift, Python and C++ tools for data collection, analysis and cardiological monitoring.
Babylon Health
Led deep-reinforcement-learning research for medical triage and active diagnosis; developed probabilistic inference, sensitivity and safety methods, simulated environments and a cross-platform inference system; published first-author research and contributed to three patent families.
05 · Publications
Aboulafia-Brakha, Buchard et al. · Psychiatry Research 358:116992 · doi:10.1016/j.psychres.2026.116992
DOI recordAboulafia-Brakha T, Buchard A, et al.. Real-world effectiveness and safety of psychedelic-assisted psychotherapy: outcomes from a large-scale compassionate-use cohort in Switzerland. Psychiatry Research 358:116992. doi:10.1016/j.psychres.2026.116992.
Moride, Hamon, Benichou, Buchard et al. · Research Square · doi:10.21203/rs.3.rs-8790368/v1
DOI recordMoride Y, Hamon Y, Benichou J, Buchard A, Grimaldi L, Abenhaim L. Validation of a high-throughput target-trial-emulation platform for drug repurposing in systemic lupus. Research Square. doi:10.21203/rs.3.rs-8790368/v1.
Buchard, Seragnoli, Sabe et al. · Research Square · doi:10.21203/rs.3.rs-8214817/v1
DOI recordBuchard A, Seragnoli F, Sabe M, et al.. Limited prognostic value of early maladaptive schemas for acute psychedelic experience and symptom improvement. Research Square. doi:10.21203/rs.3.rs-8214817/v1.
Grimaldi, Duchemin, Hamon, Buchard et al. · JAMA Network Open 7:e2432190 · doi:10.1001/jamanetworkopen.2024.32190
DOI recordGrimaldi L, Duchemin T, Hamon Y, Buchard A, Benichou J, Abenhaim L, Costedoat-Chalumeau N, Moride Y. Hydroxychloroquine and cardiovascular events in patients with systemic lupus erythematosus. JAMA Network Open 7:e2432190. doi:10.1001/jamanetworkopen.2024.32190.
Buchard, Bouvier, Prando et al. · arXiv · arXiv:2003.12828
arXiv recordBuchard A, Bouvier B, Prando G, Beard R, Livieratos M, Busbridge D, et al.. Learning medical triage from clinicians using deep Q-learning. arXiv. arXiv:2003.12828.
Albert Buchard and Jonathan G. Richens · doi:10.1007/978-3-030-64573-1_28
Jonathan G. Richens and Albert Buchard · doi:10.1007/978-3-030-64573-1_29
Dialogue flow using semantic simplexes
US 11,145,414 · Granted and active
Buchard AJT, Gourgoulias K, Zwiessele MB, Navarro AKW, Johri S
System and method for medical triage through deep Q-learning
US20210327578A1 · Published application, abandoned
Buchard A, Bouvier B, Livieratos M, Beard R, Gourgoulias K, Johri S, Prando G
Computer-implemented method and system for running inference queries with a generative model
US20210103807A1 · Published application, abandoned
Walecki R, Buchard A, Gourgoulias K, Hart C, Lomeli M, Baker A, Navarro AKW, Zwiessele M, Perov Y, Johri S
06 · Software
Creator and principal maintainer
Public · Apache-2.0 · active
Local-first structured memory and planning for people and artificial-intelligence agents, with provenance, explicit relationships and device synchronisation.
Public repository evidence; the screenshot uses repository-owned fake fixture data under Apache-2.0.
Creator and maintainer
Public · MIT · release 0.9.14
Python package for causal discovery, statistical testing, model comparison and reproducible reporting.
Newer unreleased source work is not presented as part of version 0.9.14.
Public version-0.9.14 package, documentation and repository evidence.
Doctoral research
In press
Known-truth evaluation of deep sequential models on simulated and task-based functional-magnetic-resonance-imaging data.
The work is in press. Its source repository is non-public; no repository link or internal material is exposed.
Current CV evidence and Albert’s direct status correction; no public bibliographic record has been supplied.
Creator
Public training system
Deliberate practice of psychotherapy communication skills with local speech and language processing.
Training software; no claim of clinical validation, diagnosis or treatment efficacy.
Anonymous public help view with repository source and NOTICE attribution.
Distributed PyTorch + SLURM GraphModelParser PyRetest Experiment.js Calibrator.js
07 · Education
· In progress
University of Geneva Faculty of Medicine
Causal inference in deep sequential task-fMRI models under Professor Stefan Kaiser.
Sorbonne University
French medical degree; Swiss federal recognition in 2014.
Sorbonne University
Magna cum laude; second of 60. École normale supérieure research used patch-clamp electrophysiology, two-photon imaging and immunohistology to study serotonergic modulation of cerebellar Golgi cells.
· Programme not completed
University of Geneva · Bavelier Laboratory
Hierarchical statistical learning, compositional knowledge, reinforcement learning, behaviour and neuroimaging as a Swiss National Science Foundation fellow.
· École de l’INSERM selection and scholarship
National competitive selection after application and oral examination; two-year master’s scholarship.
· Swiss National Science Foundation MD-PhD scholarship
CHF 180,000 over three years through the Lemanic Neuroscience Doctoral School.
· Stanford best poster and invited panelist
Frontier of Artificial Intelligence-Assisted Care Scientific Symposium.
2017 · Tied best-team project, CoSMo Summer School, Minneapolis.
08 · Teaching
Schema Therapy for Personality Disorders · 4.5 hours
Psychiatry Through Film: Clinical Perspectives on Mental Illness · 2 hours
Artificial Intelligence and Medicine · 2 hours · about 50 Sorbonne University students
Statistical Analysis and Machine Learning with R · about 40 hours · 20 graduate sessions · 7 students
Cognitive Artificial Intelligence Meetings · founded and organised about 20 hours at Campus Biotech
Supervised or co-supervised seven students across computer science, machine learning, medicine and data science, and psychology at the Massachusetts Institute of Technology, the University of British Columbia, University College London and the University of Geneva; four University of Geneva master’s students were co-supervised.
09 · Contact
Languages · French · English
