In order to develop scene and text comprehension tools, the SIALV laboratory is developing a number of algorithms for different tasks. These different building blocks have historically been developed in different projects, which has led to compartmentalization between projects and difficulty in sharing knowledge. We would like to develop tools to develop a stack from the data curing to the deployment of machine learning models that are reliable and efficient.

During this internship, development can occur in various place of the MLOps stack to:
• Train neural networks efficiently on multiple GPUs and multiple computational nodes for different vision and text tasks such as Generative AI (text and image), object detections, Named Entity Recognition (NER)...
• Deploy models internally to make predictions and evaluate models.
• Monitor our models.
• Store our large quantity of data.
• Maintain a model registry to track our experiments and easily deploy our best performing models.
• Visualize and annotate datasets containing multiple modalities (text, image, video, etc.) assisted by AI.
• Solutions to orchestrate, maintain and scale the technical infrastructure in our laboratory

You will also perform technological watch of MLOps tools to find the best languages, libraries or frameworks to perform the missions mentioned above.

Find more details at: https://kalisteo.cea.fr/index.php/apply-for-a-job/

In line with CEA's commitment to integrating people with disabilities, this job is open to all.