Lecture Recap | Dr. Arnaud de La Fortelle, Co-founder and CTO of Heex: From Big Data to Smart Data

On July 8, 2025, Dr. Arnaud de La Fortelle, co-founder and CTO of Heex Technologies and former Director of the Center for Robotics at Mines Paris, PSL University, delivered a keynote lecture titled "From Big Data to Smart Data: Rethinking the Edge-Cloud Data Continuum for Next-Gen AI" at GIFT. Faculty and students from the Innovation Center of Intelligent Connected Electric Vehicles, GIFT and Paris Elite Institute of Technology attended the event.

Dr. Arnaud de La Fortelle explored the conceptual evolution from "big data" to "smart data," highlighting how traditional "record-everything" approaches have become unsustainable due to the rapid development of autonomous driving, smart cities and robotic systems. "In reality, 95% of recorded data remains unused beyond 72 hours, leading to massive resource waste and carbon emissions," he noted. He emphasized the critical role of edge-cloud synergy in next-gen AI applications—while basic control systems can process data locally in real time, complex scenarios like autonomous driving still require cloud-based learning for rare events. Consequently, a dynamic, context-aware data architecture is essential.

Introducing the "smart data" concept, Dr. Fortelle advocated for event-triggered data collection. For example, a vehicle detecting a pedestrian at night would record only the relevant 10-second multi-modal clip. This method significantly reduces data volume, while preserving high-value information for machine learning. He further presented a technical framework where edge devices use embedded agents to filter and upload only task-critical data, seamlessly integrating labeling, vision, and learning processes. This "problem-driven" approach not only enhances efficiency, but also enables unified and high-quality global datasets.

"The transition from big data to smart data goes beyond technology,” he concluded. “It represents a reconstruction of the ecosystem. Only by asking why we collect data and what we aim to learn can we avoid information overload and build a greener, smarter AI future."

During the Q&A session, participants engaged in vibrant discussions on topics such as efficient historical data indexing, automated data labeling, multi-sensor triggering optimization and localization challenges for deploying global data platforms in China.

Through case studies and detailed analysis, Dr. Fortelle provided profound insights into the smart data frontiers, significantly broadening attendees' perspectives on the opportunities and challenges of this emerging era.