Memory-Augmented LLMs for Financial Text - Internship

CFM; Paris; On-site; EUR 11,000 - 16,000; Part time

Job summary

CFM in Paris invites applications for an internship focused on memory-augmented language models for financial text. You will explore maintaining a compact memory of a company’s disclosure history as new documents arrive, updating incrementally and efficiently.

The role requires strong ML/NLP skills, familiarity with LLMs, and the ability to read and implement research papers. A finance-interest is a plus, and collaboration with a driven team is expected.

Qualifications

• Currently enrolled in final year of engineering or Master’s in NLP/ML/CS.
• Solid understanding of LLMs and sequence modeling.
• Ability to read and implement research papers.

Responsibilities

• Design and compare memory-augmented strategies for updating with new documents.
• Build baselines, including a retrieval-based system.
• Evaluate representations for faithfulness to source documents and usefulness for financial prediction tasks.
• Read and implement research papers.

Skills

LLMs & sequence modelling

Read/Implement research papers

NLP fundamentals

Education

Final-year engineering or Master’s student in NLP/ML/Math/CS

Tools

Retrieval-Augmented systems

Reinforcement learning for LLMs

Job description

Select how often (in days) to receive an alert:

Memory-Augmented LLMs for Financial Text - Internship

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

How Much History Does an LLM Need? Memory-Augmented Language Models for Financial Text

Financial text about a company accumulates continuously over years. Today's language models typically process this history either by feeding in as much as fits in their context window, or by splitting it into independent chunks. Both approaches have well-known limitations, especially as history grows long and needs to stay historically consistent.

This internship explores an alternative: building a compact, incrementally updated memory of a company's disclosure history, one that is updated as new documents arrive rather than fully re-read each time.

The central question is twofold:

Does a long disclosure history help understand or predict a company's future, compared to only recent information?

If so, what is the best way to represent and maintain that history over time in a language model?

Key Responsibilities

Design and compare strategies for incrementally updating a compact memory representation as new documents arrive.

Build comparison baselines, including a retrieval-based system.

Evaluate the resulting representations both for faithfulness to the source documents and for usefulness on a downstream financial prediction task.

Read and implement research papers (a significant part of the work).

Your Skills

Required

• Final-year engineering or Master's student in NLP, Machine Learning, Applied Mathematics, or Computer Science.
• Solid understanding of LLMs and sequence modeling.
• Ability to read and implement research papers.

Appreciated

• Experience with retrieval-augmented generation, agent memory systems, or reinforcement learning for LLMs.
• Interest in financial applications (a bonus). Prior knowledge of finance is not required.
• French is a bonus.

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 .