Stage - quantitative researcher H/F

Crédit Agricole Assurances; Paris; Sur place; EUR 60 000 - 80 000; Plein temps

Résumé du poste

A leading financial institution in Paris is offering an internship focused on implementing a decision-focused learning framework for portfolio optimization. The role includes significant responsibilities in bridging mathematical methods and combinatorial optimization. Candidates are expected to show curiosity, creativity, and proactivity. Professional opportunities within the group and a financial allowance for sustainable mobility are also offered.

Prestations

Telecommuting agreement

Allowance of €600 to €700 for sustainable mobility

Professional opportunities within the group

Qualifications

• No prior professional experience required.
• Must have curiosity, creativity, and proactivity.

Responsabilités

• Implement and evaluate a decision-focused learning framework for portfolio optimization.
• Bridge gaps between mathematical methods and combinatorial optimization.
• Identify suitable algorithms for training networks.

Connaissances

Curiosity

Creativity

Proactivity

Formation

Graduate engineering school

Master in operational research

Mathematics modeling or learning vision

Description du poste

At Crédit Agricole Assurances, the Group Investment Department offers investment strategies to our various subsidiaries. It ensures their implementation and monitoring with a focus on acting in the best interest of society and our clients.

This internship is a joint project between the Quantitative Research team of Crédit Agricole Assurances and the CERMICS laboratory (École des Ponts ParisTech). The objective is to implement and evaluate a decision‑focused learning (DFL) framework applied to portfolio optimization.

Traditional Portfolio Building

1. Expected returns and dependence relationships (covariances) between assets are estimated.
2. These estimates are plugged into a mathematical model to find the optimal mix of assets according to the objective.

The main issue with this “estimate‑then‑optimize” approach is that both steps are treated as separate tasks, leading to several issues, particularly with multivariate time series data.

Internship Objective

The internship objective is to implement and evaluate a Decision‑Focused Learning (DFL) framework aiming to tackle the flaws of the traditional “estimate, then optimize” method. Instead, DFL’s core idea is to create a single, end‑to‑end system that learns to make the best possible investment decisions directly from data.

Portfolio optimization often involves making discrete (combinatorial) choices. However, the mathematical methods typically used to train neural networks work best with smooth, continuous problems, not discrete, combinatorial ones.

A key challenge will be to bridge this gap and identify the most suitable algorithm from the existing literature for training a network that includes a combinatorial optimization layer, specifically tailored to the unique uncertainties inherent in portfolio optimization.

Benefits

Professional opportunities within our Insurance entities and the Crédit Agricole group, a telecommuting agreement, a €600 to €700 allowance to promote sustainable mobility, and much more to discover!

Recruitment Process

1. An in‑person or remote interview with the manager or internship mentor.
2. A phone call with an HR representative to set up hiring arrangements (internship dates, explanation of the hiring process, agreement, compensation).

This year, over 100 interns have been welcomed at the heart of our teams. A concrete way to support young people entering the job market, in line with the commitments outlined in the Crédit Agricole Group Youth Plan.

Principles of environmental protection and social responsibility are integrated at the core of our activities.

All of our offers are open to people with disabilities.

A first professional experience is not required for the role. Curiosity, creativity and proactivity are essential.

Grande Ecole d'Ingénieur, Master en recherche opérationnelle (RO), mathématiques de modélisation ou vision apprentissage