My research on collaborative insurance and risk sharing focuses on the design and analysis of mechanisms through which individuals or institutions pool and redistribute risks. A first strand concerns the mathematical properties of risk-sharing rules, including fairness, efficiency, monotonicity, Pareto optimality and incentive-related properties. Particular attention is given to conditional mean risk sharing and to axiomatic characterizations of alternative allocation rules.

A second strand studies the behaviour of these mechanisms under dependence, heterogeneity and large-pool asymptotics. My work investigates how conditional mean risk sharing operates for independent, conditionally independent and dependent losses, how it reduces or redistributes risk, and how its properties evolve as the size of the pool increases. This includes results on tail behaviour, graphical dependence structures and asymptotic risk elimination.

A third strand concerns the design of decentralized and collaborative insurance schemes. Applications include peer-to-peer insurance, tontine and survivor-fund mechanisms, contingency funds, reinsurance and stop-loss protection, and parametric insurance with residual basis-risk sharing. These works aim at understanding how alternative forms of mutualisation can complement traditional insurance while preserving actuarial fairness and economic incentives.

My research in extreme value theory first focuses on statistical inference for rare events, with particular emphasis on the extremal index, clustering of extremes and extremal dependence. I have developed estimators for characterizing the frequency, size and persistence of extreme-event clusters, as well as inference methods for multivariate and spatial max-stable models.

A second strand concerns stochastic-process models generating extreme and heavy-tailed behaviour. My work has investigated how dynamic dependence mechanisms give rise to heavy tails, notably in stochastic recurrence and unit-root models. More recent contributions address spatial and spatio-temporal extremes through max-stable and Brown–Resnick processes, as well as geometric aspects of extremes arising in Poisson–Voronoï tessellations.

A third strand concerns actuarial applications of extreme value theory and heavy-tail methods. It includes the asymptotic analysis of large insurance losses, dependence between catastrophic risks, ruin-related problems, and efficient simulation methods for rare events. These contributions aim at quantifying the impact of extreme losses and dependence on the stability and solvency of insurance portfolios.

My research on machine learning for insurance and finance develops statistical learning methods adapted to complex risk data. A first direction focuses on interpretable machine learning for insurance, including individual claims reserving, tree-based models and methods for translating complex predictive models into more transparent generalized linear or additive representations.

A second direction concerns machine-learning methods for extreme risks, including partition-based tail-index estimation and rule extraction for rare and severe insurance losses. These works seek to combine predictive flexibility with statistical interpretability and reliable tail-risk assessment.

A third direction addresses financial applications through graph-based learning. In particular, I have studied graph neural networks and relational information for multivariate volatility forecasting and stock-movement prediction, allowing complex dependence structures between financial assets to be incorporated directly into predictive models.

My research in financial econometrics and quantitative finance focuses on the statistical modelling of financial markets, with particular emphasis on high-frequency data. A first direction concerns market microstructure, including ultra-high-frequency price dynamics, order flow, price formation and microstructural hedging errors.

A second direction addresses the estimation of latent market quantities such as efficient prices, volatility and covariation in the presence of microstructure noise and endogenous trading times. My work also considers statistical tests for the structure and dynamics of continuous-time financial processes.

A third direction concerns stochastic models for financial risk, including stochastic unit-root processes, diffusion models and credit-risk models under partial information. These works seek to connect probabilistic modelling with statistically robust inference for complex financial data.

My research in applied probability and statistics focuses on dependence modelling, stochastic processes and statistical inference. One direction concerns hierarchical copula models and composite-likelihood methods for complex multivariate dependence structures. Another investigates state-dependent stochastic processes, with emphasis on stability and long-run behaviour. More broadly, these works provide probabilistic and statistical tools for the analysis of complex dependent systems, with connections to extreme value theory, financial econometrics and risk modelling.