Internship: AI pipeline optimization

Barco

Barco designs technology that makes everyday life a little better. Seeing beyond the image, we develop sight, sound, and sharing solutions to help customers work together, share insights, and wow audiences. Healthcare is one of the key markets of Barco. For many years Barco has been contributing to improved healthcare by means of solutions in radiology, mammography, surgery, dermatology, dentistry, pathology etc.

Innovation in healthcare

The Barco Labs Healthcare team is constantly looking for new innovative solutions that push forward the state-of-the-art and can improve healthcare models. This group takes care of the entire innovation cycle: ideation and MVP definition, market evaluation and business case creation, R&D and clinical work for creation of proof of concepts and solutions, market and clinical / regulatory validation of the solutions, business model and business plan creation, up to commercial introduction and early pilot sales.

The task at hand

For this internship there are two key projects. The choice of project can be discussed based on your interests, motivation, and skills.

The first project focuses on migrating a proprietary pipeline compute step in the context of AI from its current implementation to an open-source orchestration framework, such as Ray. The existing pipeline is structured around an abstract base class in Python, where the input and output are pandas DataFrames containing metadata. The pipeline computes tasks in parallel using multiple processes, and in some configurations, multiple GPUs. The goal is to refactor this pipeline step to work seamlessly with Ray while preserving current functionality and requiring minimal changes to child classes that extend the base class.

The second project focuses on creating a dedicated AI inference framework optimized for real-time streaming data, inspired by frameworks like NVIDIA Holoscan. This project will involve designing a fully optimized, end-to-end system that efficiently handles data streaming, preprocessing, AI inference, and post-processing of results. You will identify and mitigate bottlenecks at each stage of the pipeline—such as latency in data provisioning, inefficient memory management during preprocessing, or lag in real-time inference—to ensure seamless performance, particularly in applications like real-time healthcare video analytics and efficient handling of large medical imaging tasks.

Qualifications

Generic software skills:

• Proficiency in Python programming. 

• Familiarity with object-oriented programming and software design patterns. 

• Experience working with version control systems (e.g., Git). 

• Basic understanding of working with data (e.g., pandas, data processing workflows). 

• Strong problem-solving and debugging skills. 

And interest in one or more of the following areas:

• Distributed computing frameworks (e.g., Ray, or similar). 

• Multiprocessing and parallel computing in Python. 

• Efficient low latency programming on GPU. 

• Optimizing data pipelines and performance benchmarking. 

• Refactoring and migrating legacy code into scalable, modern systems. 

Furthermore, you should be a student in a technical discipline, eligible to work at our HQ in Kortrijk Belgium, and fluent in English.