Self-driving cars are the most complex ML system that exists. Full stop.

3D object detection fusing LiDAR and camera data. Path planning combining reinforcement learning with graph algorithms. Behavior prediction using transformers. Sensor fusion across multiple modalities. All running at less than 50ms latency because lives literally depend on it.

And then simulation testing in digital twins because you can't test edge cases — "what happens when a ball bounces into the road from between parked cars" — in the real world at scale.

I find this fascinating not because I work on autonomous vehicles directly, but because the engineering challenges cascade into every other field. Real-time inference constraints. Safety-critical ML. Multi-sensor fusion. Performance-critical code in C++.

If you can build systems that work for autonomous driving, you can build ML for anything.

The companies in this space (Tesla, Waymo, and several others) are hiring aggressively for ML engineers who understand 3D vision, RL, and real-time systems. It's one of the hardest ML problems. And one of the most rewarding if you want to work on technology that will genuinely reshape how the world moves.