Research
Autonomy at Coco
Our Strategy
When we started Coco, our autonomy strategy was simple: scale a business that generates revenue and data, then use that data to build the best models. A few years later, we’re far from finished, but we’ve significantly advanced the revenue and data sides of the business. Our global fleet of hundreds of robots operates 24/7, producing tens of thousands of hours of data each month - without extra effort or dedicated teams, simply as a natural outcome of running our business at scale.
Software 2.0
Our goal is to become the default way goods move through a city. However, most robotics systems operate in structured environments - roads with lane lines, warehouses with fixed routes, or factories designed for predictability. Cities are the opposite: dense, dynamic, and unstructured. This makes the traditional robotics stack brittle and slow.
That’s why we believe any scalable, real-world autonomy must be data-driven. Andrej Karpathy describes this well in his blog post on Software 2.0. Our focus is on turning our massive, diverse dataset into end-to-end navigation policies and simulation systems that learn directly from data. With our scale and global footprint, we’re uniquely positioned to lead this next generation of autonomy.
Welcome Bolei Zhou
We recently welcomed Bolei Zhou as our Chief AI Scientist. Bolei is a world renowned AI researcher with deep expertise in computer vision, deep learning, reinforcement learning, and data driven simulation. His leadership will help us accelerate our goal of building the most capable and reliable AI systems in the world.
more to come
To learn more about Bolei and his previous research, head to his site.
