Hi! PARIS brings together an interdisciplinary community of researchers in artificial intelligence, data science, mathematics, social sciences, and management, committed to advancing rigorous, responsible, and impactful AI. Across its network, research is carried out by faculty and research teams, supported by students and postdoctoral researchers from around the world.
Organized around major research axes, Hi! PARIS builds on the academic excellence of its founding institutions and on close collaborations with industry, public stakeholders, and international partners. This ecosystem fosters innovative approaches that combine fundamental research and applied solutions to address today’s technological, economic, and societal challenges, with continuous attention to ethical, environmental, and regulatory issues.
Mathematical Foundations of AI
This research axis focuses on the core mathematical principles that underpin modern AI. Work in this area covers high dimensional probability, computational optimal transport, causal inference, implicit regularization, and scalable Bayesian inference. Research also advances optimization methods, from global to stochastic optimization, with a strong focus on performance at scale.
A key priority is massively distributed and federated learning, addressing challenges such as statistical heterogeneity, communication constraints, and privacy control. In parallel, the mathematical foundations of deep learning are explored, including training dynamics, new transformer architectures, overparameterization, uncertainty quantification, and ways to overcome the curse of dimensionality.
These efforts rely on close links between mathematics, statistics, and statistical physics to develop rigorous and reliable AI methods.
Foundation Models
This research axis focuses on the development of medium scale foundation models in natural language processing, computer graphics, and AI for sound. While building state of the art generative models is not the goal, the research agenda is ambitious and targeted.
A first priority is the design of lightweight generative models that can be trained efficiently. This includes reducing computational costs through efficient attention architectures, knowledge distillation, pruning, quantization, and improved parallelism to enable large scale training on low bandwidth GPU networks.
A second priority is robustness and fairness. Research explores data augmentation with counterfactual or debiased examples, fine tuning to align model behavior with desired values, and prompt engineering. Societal challenges are also addressed, including the detection of real versus fake content and the development of regulatory approaches.
A third priority is cross modal learning, with a focus on representation fusion across text, image, and audio, scalability of multimodal models, and evaluation metrics for quality, diversity, and coherence. This axis also explores creative AI applications in text, image, and audio generation, while addressing ethical, economic, and security issues and developing methods to detect AI generated content.
Trustworthy and Sustainable AI
This research axis focuses on building AI systems that are not only effective and mathematically rigorous, but also aligned with key societal and environmental requirements. The work spans mathematics, computer science, and social sciences.
Core research topics include privacy preserving AI, fairness, and efficiency, as well as robustness to adversarial attacks, noisy labels, and outliers. A strong emphasis is also placed on explainability, with the goal of making AI systems more transparent, reliable, and accountable in real world settings.
AI for Sciences and Engineering
This research axis applies AI to advance scientific discovery and engineering systems. In computational statistical physics, AI improves coarse graining techniques and enables advanced sampling methods for chemistry and biology, using approaches such as neural flows, variational autoencoders, and diffusion models. This work is closely connected to research in computational physics, molecular dynamics, and materials science.
Another priority is smart industries, with a focus on sustainability. Research addresses predictive maintenance, optimal control under uncertainty, and process optimization through digital twins. Smart energy is also a key area, covering building energy management, microgrid control, and energy markets, in collaboration with dedicated energy research programs.
Finally, this axis explores smart mobility, integrating AI into transportation systems to improve efficiency, safety, sustainability, and user comfort. Across all these domains, operations research plays a central role, providing the methodological foundations for large scale, reliable, and efficient AI driven solutions.
AI in Cyber-Physical Systems and Robotics
This research axis focuses on the convergence of AI with cyber-physical systems, IoT, and robotics. As these technologies reshape industries and smart cities, research addresses distributed AI for edge devices, predictive maintenance through anomaly detection, security of connected systems, and adaptive self learning systems. AI driven digital twins also play a growing role in system simulation and optimization.
Key directions include AI for real time analytics and decision making based on large sensor data streams, as well as adaptive systems that learn continuously in dynamic environments. In robotics, research explores perception at the intersection of computer vision and robotics, including the use of large scale models for 3D perception.
This axis also investigates new control paradigms based on deep learning, human robot interaction adapted to individual users, and fail safe systems that combine sensory data, decision making, and uncertainty management to enable reliable real world deployment.
AI for Society
This research axis builds strong links between AI and the human and social sciences. Here, AI is both a tool for analysis and an object of study, helping to better understand societal transformations while raising new research questions.
Key areas include AI and law, with a focus on legal AI adapted to the EU framework, algorithmic compliance, and the integration of AI into legal tools such as contracts and case law. Another priority is AI and the future of work, examining the diffusion of AI technologies, their impact on job content, skills, and labor markets, including applications in finance and the infrastructure needed to support AI.
The axis also explores NLP and social sciences, studying public discourse and media in the age of AI and social platforms. Research addresses polarization, misinformation, bot detection, fact checking, and the analysis of social trends by combining behavioral and sentiment data at scale.
AI for the Economy
This research axis explores the growing interface between AI and economics, as AI driven systems increasingly shape markets, labor dynamics, and decision making.
Key research directions include AI based market design, focusing on dynamic and strategy proof mechanisms that promote efficiency, fairness, and privacy in online platforms. Another priority is the transformation of the workforce, where economics and data science are combined to reduce labor market inefficiencies and develop practical recommender systems, in collaboration with public employment services.
The axis also studies human algorithm engagement, examining how algorithmic recommendations influence user behavior and preferences. Research further addresses interactions between algorithms, such as algorithmic collusion in energy and financial markets, as well as risks of model breakdowns. Finally, work on dark patterns investigates how AI can detect and mitigate manipulative online practices that influence consumer choices.