REEV | Stage - Explicabilité des modèles

About REEV

REEV develops advanced wearable robotic systems to support mobility for individuals with gait impairments.
Its flagship product, DREEVEN, is a motorized knee orthosis that intelligently adapts its assistance to the user’s motion (walking, sitting, standing, or climbing stairs) using onboard sensors and embedded AI.

The activity detection models at the core of DREEVEN are already developed and integrated by the AI team. They provide essential input signals to the control analytics layer, influencing torque, timing, and safety decisions.

This internship focuses on making these models interpretable and explainable, to improve clinical understanding, debugging, and system reliability.

More information: www.reev.care

Mission

The intern will work on designing, implementing, and evaluating explainability methods for the activity detection models currently used in DREEVEN.

This involves exploring both global (model-level) and local (decision-level) explanations, developing visualization tools, and assessing how explanations can support clinical and engineering workflows.

Activity

Study the architecture and outputs of existing activity detection models (already trained and validated by REEV’s AI team).
Develop explainability and visualization tools to interpret these models’ predictions over time.
Apply and compare explainability techniques such as SHAP, LIME, Integrated Gradients, Layer-wise Relevance Propagation (LRP), or Attention Visualization.
Evaluate explanations in terms of fidelity, stability, and runtime performance.
Propose methods to make explanations understandable to clinicians, testers, and control engineers.
Integrate the best-performing explainability workflow into the control analytics pipeline or a monitoring dashboard.
Document results and provide recommendations for future integration into production systems.

Required Skills

Clear written and oral communication in English.
Analytical mindset and scientific rigor.
Curiosity, honesty and initiative.
Ability to work in a multidisciplinary environment (AI, control, clinical experts).