How can researchers leverage Open Source AI with a data-centric approach

Open-source AI, particularly with open-weight models from Llama, Mistral, and DeepSeek, has shifted research focus from massive general-purpose systems toward compact, domain-specific solutions. This talk will overview the rapidly evolving open ecosystem of AI, emphasizing how researchers can adopt data-centric workflows to use AI responsibly and innovatively. By curating high-quality, rights-cleared datasets, and iteratively refining data quality, researchers can fine-tune open-weight models into transparent, reproducible, and sustainable tools. Adopting this approach democratizes AI, enhances methodological transparency, enabling labs of all sizes to build precise solutions tailored for embedded, privacy-sensitive, and energy-efficient AI applications.