Wearables & digital biomarkers
Detecting mood episodes and early warning signs from wrist-worn sensors (actigraphy, sleep, heart rate variability, electrodermal activity, skin temperature) in bipolar disorder, depression and psychosis.
Imperial College London · London, UK
Filippo Corponi MD, PhD
Consultant psychiatrist and applied machine-learning researcher.
I build and test AI that turns wearable and real-world health data into earlier, more precise mental-health care, with a focus on bipolar disorder and depression.
I am an NIHR BRC-funded Clinical Research Fellow at Imperial College London and an Honorary Consultant Psychiatrist in the NHS. I practise psychiatry and build the models myself, so I know what a tool must do for clinicians, and whether it actually does it.
- PhD in Biomedical AIUniversity of Edinburgh
- Consultant psychiatristNHS, General Adult Psychiatry
- Principal InvestigatorWellcome- and NIHR-funded projects
- 30+ publicationsWearables, ML and mood disorders
Expertise
What I work on
Machine learning for health time-series
Self-supervised pre-training, multi-task learning, deep state-space models, time-series foundation models and Bayesian modelling. Python, PyTorch, PyMC.
Real-world data at scale
Linked health records and Fitbit data from the All of Us Research Program: cohorts and outcomes defined from SNOMED and ICD-10 codes, analysed inside trusted research environments (SQL/BigQuery).
Evaluation that reflects clinical reality
Outcome definitions and labelling rules; false-positive versus false-negative costs; detection lead time against standard care; and subgroup fairness.
LLMs & AI agents in mental health
AI voice-assistant calls for remote symptom monitoring, retrieval-augmented assistants for evidence-grounded querying of health data, and tools designed to be callable by AI agents.
Clinical practice
Consultant in general adult psychiatry across community mental health, intensive home treatment and early intervention in psychosis, so models are designed around real clinical decisions.
Projects
Current projects
RISE UP
Open-source wearable AI platform for early risk detection
A four-institution team (Imperial, King’s College London, UCL, Edinburgh) building an open-source Python package and dashboard that take clinical researchers without coding experience from raw wearable data to early-risk models for depression, anxiety and psychosis. Evaluation is built in: error-cost trade-offs, lead time to detection, and fairness across age, sex and ethnicity. Co-designed with a Lived Experience Advisory Panel; the roadmap includes an LLM assistant and making the platform callable by AI agents.
Funded by Wellcome through the Mental Health Data Prize
About the Wellcome Mental Health Data PrizeHOUDINI-BD
Remote monitoring of bipolar disorder between clinic visits
A 12-month prospective study combining a wrist-worn wearable, a companion smartphone app and periodic AI voice-assistant calls to track rest-activity and speech in people with bipolar disorder, aiming to spot the early signs of a new mood episode. Sponsored by Imperial College London; data collection began in October 2026.
Funded by the NIHR Imperial Biomedical Research Centre
Study websiteMood disorders in All of Us
Wearable phenotypes at population scale
Using linked health records and Fitbit data from the US All of Us Research Program to characterise mood disorders between episodes: rest-activity patterns and their modulation by photoperiod, and cognitive functioning and its relationship with rest-activity profiles.
Two first-author preprints, medRxiv 2026
TIMEBASE
Open clinical wearable dataset and benchmark
An open dataset from Hospital Clínic de Barcelona: 157 participants with bipolar disorder, depression and healthy controls; 265 recordings of about 48 hours across illness states, with clinician-rated scales and medication. Benchmark tasks of graded difficulty with subject-level splits, gated by a de-identification check.
PhysioNet release in preparation
Study protocol (BJPsych Open)Research
Selected publications
- Beyond the hype: why artificial intelligence is not a shortcut to solving mental illness BJPsych Open, 2026
- Damped physical activity and unstable sleep: Fitbit-derived rest-activity phenotypes in inter-episode mood disorders medRxiv (preprint), 2026
- Automated mood disorder symptoms monitoring from multivariate time-series sensory data: getting the full picture beyond a single number Translational Psychiatry, 2024
- Wearable data from subjects playing Super Mario, taking university exams, or performing physical exercise help detect acute mood disorder episodes via self-supervised learning JMIR mHealth and uHealth, 2024
- A Bayesian analysis of heart rate variability changes over acute episodes of bipolar disorder npj Mental Health Research, 2024
All publications · also on Google Scholar
Background
Experience
-
Nov 2025 – present
NIHR BRC-funded Clinical Research Fellow
Imperial College London, Department of Brain Sciences
Principal Investigator of HOUDINI-BD and team lead of RISE UP. Analyses of linked health records and wearable data in the All of Us Research Program to build tools for remote monitoring and early detection.
-
Jul 2026 – present
Honorary Consultant in General Adult Psychiatry
West London NHS Trust, Early Intervention in Psychosis
Diagnostic assessment, risk formulation and treatment planning for young adults with early psychosis and co-occurring mood, neurodevelopmental and substance-use conditions.
-
May 2022 – Oct 2025
Locum Consultant in General Adult Psychiatry
NHS Lothian, Community Mental Health, Edinburgh
Consultant responsibility for a community general-adult caseload across the care pathway, from referral triage and diagnosis to treatment and risk management; chaired multidisciplinary team meetings.
-
Jan 2022 – Jan 2023
Data Scientist
Hospital Clínic de Barcelona, University of Barcelona (remote)
Built the wearable-data preprocessing pipeline reused across the group’s later studies, and led ML development in PyTorch to detect acute mood episodes and estimate individual items of clinician-rated symptom scales, using multi-task learning and self-supervised pre-training. Funded by the Baszucki Brain Research Fund (Milken Institute); led to three first-author papers and a NeurIPS 2022 workshop paper.
-
2016 – 2020
Psychiatry specialist training
University of Bologna · Hospital Clínic de Barcelona · King’s College London
Specialist training in psychiatry at the University of Bologna, including a year in the Bipolar and Depressive Disorders Unit at Hospital Clínic de Barcelona, and a visiting research post in medical statistics and machine learning at King’s College London (IoPPN).
Education
- PhD and MSc by Research (Distinction) in Biomedical AI University of Edinburgh, School of Informatics, 2020–2024. UKRI-funded. Thesis: Clinically-interpretable and large-scale machine learning to monitor mood disorders with wearables.
- Specialist qualification in Psychiatry University of Bologna, 2020
- Medical Degree and Diploma in Medical Sciences (both with Distinction) University of Pisa and Scuola Superiore Sant’Anna, 2009–2017
Peer review, editing & teaching
- Trainee Editor, BJPsych Open Royal College of Psychiatrists, 2026–present
- Reviewer UKRI Medical Research Council fellowship applications; expert reviewer for a Delphi consensus guideline on bipolar disorder; peer reviewer for Translational Psychiatry, Acta Psychiatrica Scandinavica, Journal of Affective Disorders, PLOS ONE and BJPsych Open
- Teaching & supervision MSc tutor at Imperial and Edinburgh; co-supervised four MSc students and two research assistants
- Honorary Research Fellow Centre for Clinical Brain Sciences, University of Edinburgh, 2025–2027
Contact
Get in touch
I’m keen to hear from companies and research teams working on AI, wearables and digital health for mental health. Get in touch about advisory work, collaborations, or roles.
Based in London, UK · English and Italian · CV available on request