Background
Developing robust algorithms for self-driving cars requires sourcing event data from over 10 billion hours of recorded driving time. But even with 10 billion-plus hours, it can be challenging to capture rare yet critical edge cases like weather events or collisions at scale. Luckily, these rare events can often be simulated ...