Jérémy Cohen on Smarter Data for Autonomous Systems

A Leading Voice in Autonomy

In the fast-moving world of autonomous systems, few voices manage to be both technically rigorous and widely influential, Jérémy Cohen is one of them. As the founder of Think Autonomous, he has built a platform that reaches thousands of engineers daily through more than 1,000 private emails and over 16 advanced online courses. Recognized among the “40 Under 40 Innovators” and awarded Top Global Business of the Year in EdTech, Cohen blends an educator’s clarity with a deep practitioner’s insight.

But his influence isn’t just about numbers, it’s about perspective. Across nearly a hundred blog posts, he has often questioned the “more data is always better” mindset, highlighting instead the value of smarter approaches that prioritize relevance over raw volume. Whether breaking down the nuances of LiDAR vs. RADAR, exploring sensor fusion strategies, or mapping the shift from manual data collection to edge intelligence, his work reflects a constant search for ways to make engineers faster, sharper, and more effective at solving the problems that matter.

In his latest blog post, that perspective takes a direct look at one of the most persistent challenges in the industry and at how Heex Technologies fits into the solution.

From Problem to Solution

In this recent piece, Cohen focuses on a principle he has increasingly emphasized: autonomy doesn’t progress by recording everything, but by capturing only the moments that truly matter. He frames this as a move toward event-first thinking, a way to tackle rare “long-tail” scenarios more efficiently, reduce operational overhead, and enable faster iteration. From perception and safety to system architecture, his reasoning is clear: data workflows should serve decision-making, not overwhelm it.

In “How to Stop Recording 100% of What Self-Driving Cars See (Introduction to Event Driven Automotive Data Processing)”, he applies this principle to one of the industry’s biggest bottlenecks: the inefficiency of “record-everything” pipelines. He gives concrete examples: missed pedestrian detections, fragmented datasets causing conflicting decisions, delayed visualizations that slow debugging, and the operational burden of physically swapping SSDs.

After reviewing existing approaches, he shifts to what he presents as the most effective alternative: an event-driven model that filters and captures critical moments in real time through smart triggers. The core of the article is dedicated to exploring this approach, with Heex Technologies serving as the central example of how such a model can directly address these challenges and enable teams to work faster, leaner, and more intelligently.

An Article Worth Reading

Cohen’s piece offers a clear problem–solution narrative that connects strategic thinking with practical application. For anyone involved in R&D, validation, or fleet operations, it’s a concise but impactful look at how smarter data practices can reshape the way teams work.

We value the depth with which he examined our work and the precision with which he placed it within the broader context of data processing. His perspective makes this article worth reading not just to understand our approach, but also to explore the wider evolution of automotive data management.

If you’re interested in his full perspective, you can read the complete article here: How to Stop Recording 100% of What Self-Driving Cars See.