Research Internship on Memory Mechanisms for Self-Evolving LLMs

In a nutshell

Explore how persistent memory can make large language models smarter over time by researching and prototyping memory mechanisms for self-evolving LLMs within Toyota’s AI Core Research team

Company Information

TOYOTA is one of the world’s largest automobile manufacturers and a leading global corporation. Founded in 1937. Toyota now sells vehicles in 170 countries and employs over 350.000 people. Based in Brussels, Belgium, and staffed by 2.700 people and more than 60 nationalities, Toyota Motor Europe (TME) handles the wholesale marketing of Toyota and Lexus vehicles, parts & accessories, and manages Toyota’s European R&D, manufacturing, and engineering operations.

We are excited to be transforming into a mobility company focusing on bringing even more happiness to our customers through various mobility solutions.

Your team and your responsibilities

Team/division description

The Advanced Technology division is responsible for creating cutting edge technology breakthroughs which meet the future challenges of the automotive industry and brings a smile to our customers’ face. By combining in-house research activities with a wide research network across Europe, we conceive, incubate and test novel concepts in virtual and non-virtual environments before they are handed downstream towards future product development. The position is with AI Core Research team which is in charge of performing cutting edge research on machine learning and computer vision. This internship will be supervised by Dr. Rahaf Aljundi.

Your project

Large language models achieve strong generalization through scale and test-time computation, yet they typically treat each inference as an isolated event. This proposal investigates how persistent memory can complement test-time scaling, enabling models to accumulate, organize, and reuse experience across tasks. We focus on two research questions:[RQ1] Under a fixed inference-time compute budget, how should compute be allocated across model size, memory size, and conventional test-time scaling strategies (e.g., sampling or chain-of-thought depth), to achieve efficiency, performance and learning.

[RQ2] How this memory should be designed and modified? (memory representations + memory structure)

Your Profile

• You are fluent in English (TME’s business language)
• Strong background on machine learning.
• Preferable experience with language models.
• Experience with deep learning frameworks.
• Ability to understand research papers, implement and propose ideas.
• You are fluent in English (TME’s business language)

Starting date

•  Before 30 March

Duration

•  6 months

Confidentiality

Due to business requirement, not all performed projects can be reflected in the internship report. This issue needs to be discussed with candidate/school in advance.

It is mandatory that applicants are students for the entire period of the internship.

We can only accept students with an EEA or Swiss citizenship (not requiring a work permit or visa to be eligible to perform an internship in Belgium)

We offer

We offer a dynamic multicultural work environment with broad opportunities to learn every day and exciting career pathways that help you explore different disciplines or areas of expertise.

Your benefits

• Flexible working hours and hybrid working (homeworking, 10 days from abroad);
• Company restaurant – Warm dishes, salads and desserts daily, including Japanese and traditional Belgian specialties, as well as a vegan/vegetarian option;
• Sport facilities – Tennis, squash, volleyball, football and softball pitches are available at various Toyota sites;
• Team building activities and other personnel related events;
• Programs to support your physical and social well-being.

A truly diverse, equitable and inclusive workplace

At Toyota Motor Europe, we are committed to providing equal employment opportunity to all. All our recruitment decisions are based on the individual and their qualifications along with the job requirements. The selection process is without regard to nationality, sexual orientation, gender identity or expression, age, religion, ethnicity, disability or other personal characteristics. We do not tolerate harassment or discrimination as our individual experiences and overall diversity is our key strength. Together we have the power to make a positive change and go beyond the ordinary.

If you need any specific consideration for the interview process due to disability or any other reason, please let the recruiter know.