Internship - Outcome-Guided Financial Reasoning to Predict the Markets

Paris, 75, FR

ABOUT CFM

Founded in 1991, we are a global quantitative and systematic asset management firm applying a scientific approach to finance to develop alternative investment strategies that create value for our clients.

We value innovation, dedication, collaboration, and the ability to make an impact. Together, we create a stimulating environment for talented and passionate experts in research, technology, and business to explore new ideas and challenge existing assumptions.

Your Role

Financial analysts build conviction through structured reasoning: tracing a causal narrative from disclosed information to an eventual outcome. This rationale is much easier to produce once the outcome (deal completion, market return, analyst revision) is already known.

The central project of your internship is to generate financial reasoning conditioned on an observed outcome ("hindsight-guided" reasoning), then to post-train a language model (via SFT, DPO, GRPO or related methods) so that it reproduces the quality of that reasoning without access to the outcome at inference time.

The project focuses on:

• Constructing hindsight-guided reasoning from source documents.
• Building reward models grounded in market feedback, analyst revisions and other signals.
• Training and evaluating outcome-guided reasoning models against direct prediction.

Key Responsibilities

• Get familiar with large financial text datasets and relevant targets (market returns, deal outcomes, analyst revisions).
• Generate hindsight-guided summaries and reasoning from source documents.
• Define and combine reward signals: accuracy against outcomes or market feedback, precision/recall of generated reasoning facts against hindsight-guided summaries or other datasets, and LLM-as-judge factuality and groundedness scores.
• Build the training infrastructure and train off-policy and on-policy methods on open-source LLMs, using efficient and distributed training techniques.
• Benchmark and analyze all methods (SFT, SFT with hindsight, GRPO, etc.) on out-of-sample accuracy, reward values and downstream sentiment.
• Document your methodology, results and limitations, and present them to the team.

Your Skills

• Engineering school and/or Master's student (M2 or gap year welcome), ideally with a focus on NLP or ML.
• Experience training models in PyTorch.
• Comfort working with large volumes of text.
• Ability to communicate results and limitations clearly and concisely.
• Comfortable working in English.

Nice to have

• Experience post-training LLMs, including reward modeling or preference/RL-based fine-tuning.
• Experience writing efficient PyTorch code, especially in distributed setups.
• Experience with AWS and cloud computing.
• Knowledge of finance (no prior background required; the team provides a research library and relevant literature).
• French.

EQUAL OPPORTUNITIES STATEMENT

We are continuously striving to be an equal opportunity employer and we prohibit any discrimination based on sex, disability, origin, sexual orientation, gender identity, age, race, or religion. We believe that our diversity, breadth of experience, and multiple points of view are among the leading factors in our success.

CFM is a signatory of the Women Empowerment Principles.