CEA-List

About the Company

Digital technology at the heart of our research Based in Paris-Saclay and Grenoble, the CEA-List is an institute of the CEA's Technological Research Division. Dedicated to smart digital systems, it contributes to the development of business competitiveness through the technology development and transfer. The expertise and skills developed by CEA-List's 1,000 research engineers and technicians enable our institute to support more than 200 French and foreign companies in their technological developments every year. Our collaborative projects are based on more than 700 patents, the result of research carried out by our teams, but also on a set of 21 development environments and 9 technological platforms. The quality of our research partnerships has earned us the CEA LIST Carnot Institute label for the 4th time in 2020. Our differentiating assets - Scientific and technological excellence illustrated by scientific publications in top-tier international conferences and journals (300 publications per year); - Industrial valorisation of our technologies thanks to a large portfolio of generic technologies (708 patents and 162 active licences); - effective national and international ecosystems, involving our industrial, academic and institutional partners (40 joint R&D labs and programmes); - our international outreach, with more than 100 new European projects every year and collaborations with world-class foreign academic laboratories. Find all our job offers on: www.emploi.cea.fr

Listed Jobs

Company Name

CEA-List

Job Title

Design of a Reinforcement Learning–Driven Scheduler for Efficient and Frugal Container Orchestration

Job Description

**Job Title** Design of a Reinforcement Learning–Driven Scheduler for Efficient and Frugal Container Orchestration **Role Summary** Master’s‑level internship focused on creating, training, and evaluating a reinforcement learning (RL) agent that schedules containers within distributed orchestration frameworks (e.g., Kubernetes). The role involves research, algorithm design, feature engineering, and experimental validation to improve performance, energy efficiency, and resource utilization in cloud/edge environments. **Expectations** - Final‑year Master’s or Engineering student in Computer Science, Artificial Intelligence, or related field. - Strong programming abilities (Python preferred). - Demonstrated interest in machine learning and distributed systems. - Curiosity, creativity, and robust problem‑solving skills. **Key Responsibilities** 1. Review and understand the existing orchestration framework used by the team. 2. Conduct a literature review of current RL‑based scheduling methods for cloud/distributed systems. 3. Design the architecture of an RL scheduler and select appropriate RL algorithms (e.g., DQN, PPO, actor‑critic). 4. Implement the scheduler and integrate it with the orchestration environment. 5. Develop a feature extraction module to characterize container behavior (CPU, memory, I/O, network patterns). 6. Train the RL agent, tune hyperparameters, and manage state representation. 7. Design and run experiments to benchmark the scheduler against baseline methods. 8. Analyze results, document findings, and propose performance improvements. **Required Skills** - Programming: Python (plus libraries such as NumPy, Pandas, TensorFlow/PyTorch). - Machine Learning: familiarity with reinforcement learning concepts and libraries. - Distributed Systems: basic knowledge of container orchestration (Kubernetes, Docker Swarm). - Data analysis and visualization. - Strong analytical and problem‑solving ability. - Good written and verbal communication skills. **Required Education & Certifications** - Current enrolment in a Master’s (Bac+5) or equivalent engineering program in Computer Science, Artificial Intelligence, or related discipline. - No specific certifications required. ---

Palaiseau, France

Hybrid

04-11-2025