Stage - Deep Learning Scientist

About the role

At Safran.AI, you will join a world-class team dedicated to applying cutting-edge artificial intelligence to complex real-world data, including high-resolution satellite imagery, full-motion video (FMV), and acoustic signals. As a Deep Learning Scientist within the IMINT (IMagery INTelligence) detector team, you will develop algorithms that automatically detect and classify objects critical to intelligence, defense, and aerospace applications. Since Safran's integration in September 2024, Safran.AI has also extended these AI capabilities to Industry 4.0 domains, such as automated image analysis for inspecting critical industrial parts. You will work alongside 250 highly skilled collaborators in a multidisciplinary environment that demands excellence and fosters innovation.

The IMINT detector team consists of approximately twenty experts, the majority of whom are Deep Learning Scientists, organized into three sub-teams of five to seven members each. You will focus on developing or enhancing specific object detectors for satellite imagery, including optical, infrared, and synthetic aperture radar (SAR) modalities. To support this work, the team relies on a robust internal technology stack: a generic AI Platform developed by AI Engineering for all product lines, and an IMINT-specific toolset created by the IMINT team itself. You will leverage our internal AI framework, which incorporates state-of-the-art technologies to maximize algorithm performance and uses MLOps tools to industrialize the training process.

What you'll do

• Develop and train deep learning models that detect and classify objects in satellite images across multiple imaging modalities.
• Engage in the complete algorithm lifecycle, from understanding user requirements and designing experiments to packaging and deploying detectors in operational settings.
• Contribute to the design of experimental protocols that validate detector accuracy, robustness, and generalization across diverse environments and conditions.
• Use and refine the internal AI Platform and MLOps tools to streamline data versioning, model training pipelines, and deployment workflows for the IMINT product line.
• Analyze model failures systematically to guide improvements in network architecture, training strategies, and data augmentation techniques.
• Collaborate with AI engineers to integrate algorithms into scalable, reliable, and industrial-grade software systems.
• Define and refine evaluation metrics and benchmark suites to ensure continuous performance improvements in new detector releases.
• Document technical methodologies and results clearly to support internal reviews and cross-team knowledge sharing.
• Support the transition from prototype experimentation to production by preparing models for deployment in customer environments.
• Act as an active technical contributor within a multidisciplinary team, exchanging ideas with specialists in radar, electro-optics, and signal processing.

Requirements

• Currently enrolled in a degree program related to Software Engineering, Machine Learning, Computer Vision, or a closely aligned field.
• Strong theoretical and practical understanding of deep learning techniques for imagery, including convolutional neural networks and modern object detection architectures.
• Solid proficiency in Python and adherence to strong software engineering practices such as version control, testing, and modular design.
• Demonstrated rigor in debugging models and creating reproducible experiments using internal tooling and frameworks.
• Ability to thrive in a fast-paced environment where priorities adapt to operational needs and evolving project demands.
• Openness to feedback and eagerness to learn from senior scientists and engineers across disciplines.
• Effective communication skills to explain technical trade-offs and collaborate efficiently with multidisciplinary teams.
• Curiosity and motivation to expand your expertise beyond deep learning into related areas such as data curation, deployment strategies, and system integration.

Experience with satellite or remote sensing datasets is a significant advantage, as it provides deeper insight into domain-specific challenges. Familiarity with MLOps frameworks and cloud deployment patterns enhances your ability to scale experiments and streamline production workflows. Knowledge of radar or multi-sensor fusion techniques can further strengthen your contributions to the IMINT product line.

This internship is based in Paris and follows standard academic scheduling. The role is primarily on-site, and travel is generally not required. Safran.AI offers competitive internship benefits, including a Swile card subsidy, transportation support, and access to well-being services. Candidates requiring visa support may be considered under company policy and local regulations.