From the Data and AI Systems Laboratory (DASLab) at Harvard University, Greek researcher Stratos Idreos is leading a transformation in the approach to artificial intelligence, advocating for a systemic and academic vision. In a tech sector marked by investor pressure and rapid AI infrastructure expansion, Idreos emphasizes the need to rethink the systems that support this technology to make it truly accessible to individuals and businesses. According to Idreos, the key lies in developing systems specifically designed for AI, capable of eliminating the complexity that currently hinders its adoption. He recalls that historically, major technological advances have emerged by simplifying complex problems, as seen with operating systems, databases, or cloud computing. Currently, AI involves enormous complexity: models, data, people, and multiple moving components. However, current models are insufficient to solve real-world problems at scale, as they require complex and costly infrastructure, inaccessible to most. DASLab is working on systems that allow for the rapid and easy creation of customized AI models, democratizing its use as SQL did for databases. The goal is for anyone to build AI solutions without being an engineer or worrying about data or model management, removing technical barriers and accelerating innovation. Idreos advocates for self-designed systems that can automatically adapt to the context, data, and user needs, generating optimized algorithms and models for each case. These solutions, already tested in image AI and large language models, promise to be up to a thousand times faster than current systems and would enable businesses and individuals to efficiently solve specific problems. To scale this approach, Idreos and his team have created a startup that demonstrates the potential of personalized intelligence in concrete use cases. He believes that while technological infrastructure will continue to grow, the real advancement will be making AI accessible and useful for everyone, avoiding dependence on large corporations and promoting user independence. Regarding data management, Idreos advocates for the importance of each person or entity being able to create and control their own models with private data, without relying on centralized platforms. This would allow for addressing local or specific problems, such as rare diseases, with models trained even on personal devices, provided the right tools are available. On data sovereignty, he acknowledges European concerns about protecting intellectual property and privacy but also highlights the value of data sharing to generate collective benefits, especially in sectors like health. He believes that decisions about data use and protection should be made by those who deeply understand society, not just technologists. Idreos also explores alternatives to the JPEG format, proposing self-designed systems that optimize image storage and processing according to the use case, achieving solutions up to ten times faster in AI applications. Finally, he warns about hardware challenges and the need to rethink the production and design of AI equipment, an area where public intervention could be key to driving disruptive innovations that the private market cannot undertake alone.