Ziyad Ghezal

Led the second line of defence (LoD2) initial review of Crédit Logement's machine-learning-based residential property valuation model.

• Reviewed the valuation model methodology, data sources and backtesting results, including benchmarking across model versions.
• Assessed the integration of physical climate risk indicators and the projection logic across scenarios and time horizons.
• Evaluated the prudent value computation against CRR requirements (Articles 208–229).
• Drafted the formal documentation request to the first line of defence and the framework for the final validation opinion.

• Rationalisation and optimisation of production processes
• Conception and design of Data Quality solution to ensure the reliability of data feeding the calculation engines while supporting maintainability and process automation.
• Integrating physical and transition risks in stress testing framework by enriching production processes with climate data
• Add-hoc analysis to propagate stress on collaterals, impact assessment on EAD/RWA/provisions of climate risk.

• Ownership of the functional design and technology choices for the target solution (acting as an internal product owner / business sponsor on behalf of Risk), including the selection of the architecture patterns and tooling.
• Design of a Basel IV / CRR3 impact simulator for credit and counterparty risk portfolios: simulation of the new regulatory approaches (,
  estimation of RWA impacts, expected losses and capital buffers
• Operated within an Agile delivery framework, contributing to the definition of the target architecture, including the calculation grid and integration with risk systems.
• Structuring of calculation runs (multi-scenario, regulatory uncertainty analysis) under strong volume and performance constraints.
• Coordination with Market / Counterparty Risk and Finance teams to ensure methodological alignment and industrialisation.
• Add-hoc analysis: Turnover analysis under BIV, CRM impact assessment, transition risk screening by sector (NACE) and physical risk mapping by geography...

• Independent validation of PD and LGD models on Large Corporate portfolio, covering conceptual soundness, data quality, calibration and discriminatory power, back-testing, stability and sensitivity analyses, implementation review and formal reporting to Model Risk governance.

• End-to-end development of the Reference Data Set (RDS) Large Corporate for LGD modeling, including the reconstruction of cash-flow histories and exposures at facility level. Built the observed LGD by tracking changes in outstanding amounts over the life of defaulted facilities, using a dedicated reconciliation algorithm to align recoveries, write-offs and workouts with the original exposures.
• Design and calibration of LGD models for Large Corporates under TTC (Through-the-Cycle) approache: definition of segmentation, selection of risk drivers, parameter estimation, risk calibration, MOC estimation. Performed in SAS and Python.
• Performed detailed impact analyses on RWA, expected loss and provisioning, and produced formal model documentation (methodology, assumptions, limitations, governance).

• Took part in a Quantitative Impact Study (QIS) on the treatment of NDOD (Non-Defaulted Obligors) within the bank’s IRB framework, with a focus on both Retail and Corporate portfolios.
• Performed an independent review and validation of the work done by risk modelling teams on the incorporation of NDOD in risk parameters and capital calculations (PD, LGD, RWA).
• Challenged methodological choices, data sources, segmentation and assumptions, and verified the consistency of NDOD treatment across portfolios and business lines.
• Replicated key metrics and carried out quantitative checks (back-of-the-envelope benchmarking, sensitivity analyses) to assess the robustness and risk impact of the NDOD framework.
• Assessed regulatory compliance of the NDOD implementation with internal standards and supervisor expectations, and issued findings and recommendations presented to Model Risk / Internal Audit committees.

• Worked within the risk modelling team in charge of quantitative frameworks supporting Solvency II capital requirements and portfolio risk measurement for the credit insurance book.
• Solvency II: contributed to the design and enhancement of models used for the Solvency Capital Requirement (SCR) and technical provisions, with a focus on default and claims risk on insured portfolios (by country, sector, rating, buyer).
• Claims prediction: developed and calibrated models to predict claims frequency and severity, using historical claims, exposure and macroeconomic drivers.
• Unexpected loss modelling: contributed to the development of Unexpected Loss (UL) metrics at portfolio and sub-portfolio level, combining probability of default, exposure at default and loss given default; analysed diversification effects and concentration risk for management and risk committees. Developed in C++.
• Produced quantitative analyses and dashboards to support Risk, Underwriting and Actuarial teams, and participated in the documentation and governance processes required by Solvency II model validation (internal and external reviewers).

• Design and calibration of PD/LGD models (PIT and TTC approaches for IFRS9 and Basel purposes), including parameter estimation and backtesting.
• Multi-scenario Credit VaR simulations (Monte Carlo) and impact analysis at portfolio level.
• Impact analysis, formal documentation and handover to production of validated models.
• Automation of data pipelines (data quality, traceability, reproducibility) to support risk reporting.
• Participation in internal and regulatory stress tests (methodological design, execution and results presentation).
• Participation to key regulatory exercises: ICAAP, TRIM, EBA stress tests, liquidity ratio.
• Further developed expertise in R, Matlab and Python for quantitative modelling and analysis.

• Developed a Java-based validation tool for SWIFT / T2S post-trade messages, as part of the bank’s migration to the TARGET2-Securities (T2S) platform.
• Implemented XML/XSD parsing and business rule checks to control the integrity and consistency of settlement and custody messages.
• Optimised processing logic to handle high message volumes with strong requirements on robustness, stability and error handling.
• Worked in a Java/J2EE environment in close collaboration with project and production teams, contributing to the industrialisation of post-trade workflows.