Description

You will develop a framework for merging pretrained models, focusing on distillation objectives and evaluations to improve model performance in 3D perception tasks.

Requirements

• •Master's degree in Computer Science or related field
• •Experience with deep learning frameworks
• •Excellent programming skills in Python
• •Knowledge of perception algorithms is a plus
• •Goal-oriented and structured work style
• •Very good English skills
• •Enrollment at university

Tasks

• •Develop a multi-teacher knowledge distillation framework.
• •Merge multiple pretrained models into a compact backbone.
• •Focus on task-agnostic distillation objectives.
• •Align heterogeneous feature spaces to prevent negative transfer.
• •Integrate supervised teachers trained on unknown tasks.
• •Aim for stronger, general representations with reduced model capacity.
• •Ensure compatibility with existing perception stacks.
• •Conduct structured evaluations using public datasets.
• •Utilize internal real-world data for practical results.

Tools & Technologies