Student project: Optimizing Open-set Recognition Methods for Reliable Real-world AI Systems

OSR is critical for robust and safe real-world AI systems in fields like: autonomous driving, medical diagnosis, security and surveillance, manufacturing and quality control. This project involves methodology optimization for generic OSR tasks.

What you will do

We are seeking an intern with a strong background in machine learning and mathematics to join our team and contribute to the development of robust OSR methodologies. This internship offers a unique opportunity to work on a cutting-edge research problem that is critical for ensuring safe and reliable AI deployment in real-world applications such as autonomous driving, medical diagnosis, security and surveillance, and industrial quality control. You will collaborate with a multidisciplinary team to design, implement, and evaluate OSR techniques that enable AI systems to recognize when an input does not belong to any known category.

In short, the internship involves:

• Methodology Optimization:
  Explore and refine OSR algorithms to improve their ability to detect unknown classes while maintaining high performance on known data. This includes investigating both classical machine learning and modern deep learning approaches.
• Feature Analysis and Representation Learning:
  Study the behavior of feature spaces under open-set conditions and develop representations that enhance class separability and uncertainty estimation.
• Model Development and Evaluation:
  Design, train, and test OSR models across diverse datasets, benchmarking against state-of-the-art methods to assess generalization, robustness, and scalability.
• Algorithm Efficiency and Reliability:
  Optimize OSR models for computational efficiency and stability, ensuring that recognition performance remains consistent under varying data conditions and noise.
• Documentation and Reporting:
  Maintain clear and structured documentation of methodologies, experimental setups, and results. Present progress and findings to the project team and contribute to potential research publications.
• Collaboration and Communication:
  Engage with senior researchers and developers to integrate feedback, discuss technical challenges, and iteratively improve the OSR framework based on experimental outcomes.

The internship work and activities will be organized with a scrum-like methodology: you will maintain the backlog in coordination with your mentors. You will select prioritized tasks from the backlog, and you will tackle and evaluate them on a biweekly basis. At the end of each biweekly iteration, you will showcase the progress made and will reflect on insights and improvements to focus on. Additional stakeholders may take part in the showcases to get better feedback on the study.

What we do for you

• We have a diverse team of experts both from the technical and biomedical sides to supervise and support you.
• We have a challenging problem where you have the freedom to help develop it in a specific direction.
• You will join the Digital Twin team of OnePlanet, which employs state-of-the-art knowledge on machine learning for precision medicine.
• You will be able to exchange views and knowledge with the OnePlanet and Imec community of experts and scientists, widening your professional network.
• At OnePlanet we embrace diversity and thus give equal opportunities to intern candidates with diverse backgrounds.

Who you are

• MSc student enrolled in Mathematics, Computer Science, Electrical/Biomedical Engineering, or a related field.
• Machine Learning Knowledge: Understanding of most machine learning concepts, with hands-on experience in model building and evaluation (e.g., classification models, performance evaluation, etc).
• Programming Skills: Proficiency in Python and machine learning libraries such as scikit-learn, TensorFlow, or PyTorch.
• Analytical Skills: Ability to analyze large datasets and extract meaningful insights for model improvement.
• Plus – experience with computer vision.
• Plus – Leverage cloud platforms such as Microsoft Azure or HPC environments to manage large datasets and accelerate model training and optimization.

Interested

Does this position sound like an interesting next step in your career at imec? Don’t hesitate to submit your application by clicking on ‘APPLY NOW’.
Should you have more questions about the job, you can contact jobs@imec.nl.