New Way Now: Nuro is creating a safe path to a driverless future with Google Cloud
33๐ฆ๐๐บ๐บ๐ฎ๐ฟ๐: Andrew Clare, Chief Technology Officer of Nuro, shares how the innovative robotics company is using Google Cloud to develop the Nuro Driver โ its cutting-edge driverless autonomous technology. Leveraging Google Cloud, Nuro can utilize all of its sensor data at scale for training and improving the state-of-the-art AI models that enable the system to understand and safely navigate public roads.
๐๐ต๐ฎ๐น๐น๐ฒ๐ป๐ด๐ฒ: To support its mission of deploying safe and reliable technology, Nuro needed the right AI infrastructure to create an efficient data cycle and data engine capable of powering its AI and scaling the Nuro Driver across multiple cities and geographies.
๐ฆ๐ผ๐น๐๐๐ถ๐ผ๐ป: Google Cloudโs AI-optimized infrastructure is built to deliver global scale and performance for serving billions of users, providing the hardware and software capabilities Nuro needs to scale its autonomy stack. Using Google Cloud TPUs, Nuro can now process petabytes worth of real-world driving data and train AI models for producing city-scale maps and detecting obstacles on the road twice as fast without incremental cost. The company also leverages Google Kubernetes Engine to run large-scale simulations for validating the safety and performance of AI models before release.
๐ฅ๐ฒ๐๐๐น๐๐: Now, Nuro can fuel rapid iteration cycles to continuously train and improve its autonomy stack to deliver AI-first driverless technology that can operate and navigate any kind of vehicle โ from passenger cars to semi-trucks โ safely on the road.
๐๐ป๐๐ฒ๐ฟ๐๐ถ๐ฒ๐ ๐ต๐ถ๐ด๐ต๐น๐ถ๐ด๐ต๐๐ ๐ฎ๐ป๐ฑ ๐ธ๐ฒ๐ ๐๐ฎ๐ธ๐ฒ๐ฎ๐๐ฎ๐๐ ๐ณ๐ฟ๐ผ๐บ ๐๐ป๐ฑ๐ฟ๐ฒ๐ ๐๐น๐ฎ๐ฟ๐ฒ, ๐๐ต๐ถ๐ฒ๐ณ ๐ง๐ฒ๐ฐ๐ต๐ป๐ผ๐น๐ผ๐ด๐ ๐ข๐ณ๐ณ๐ถ๐ฐ๐ฒ๐ฟ, ๐ก๐๐ฟ๐ผ:
โ โWe wanted to move to a more modern architecture. We moved from doing training on Google Cloud GPUs over to TPU v5, and what that enabled us to do was to accelerate the training time of that model by 2x without any incremental cost.โ
โ โWe've been a proud partner of Google Cloud for many years now. It all starts with the massive amount of data that Nuro has to process. We have a large in-house fleet of vehicles that are always gathering sensor data from driving on public roads that produces petabytes worth of data that we need to manage and then utilize for all of the training of our various AI models.โ
โ โWe'll take hundreds of thousands of miles of driving data, and we'll actually use that to run a large-scale simulation on GKE. It allows us to do really rapid iteration cycles to continue to improve our autonomy stack and also ensures that when we are ready to go driverless with a new release, we are really confident in the safety of that release.โ
โ โI'm super excited about what Nuro is building. I think in the next five years we're going to go from a small quantity of vehicles with L4 autonomous technology to millions. It's enabled by all the recent advancements in AI. It's enabled by the partnership that we have with Google Cloud.โ
๐๐ผ๐ผ๐ด๐น๐ฒ ๐๐น๐ผ๐๐ฑ ๐ฝ๐ฟ๐ผ๐ฑ๐๐ฐ๐๐ ๐๐๐ฒ๐ฑ: AI Infrastructure, Google Kubernetes Engine, Vertex AI
๐๐ฒ๐ฎ๐ฟ๐ป ๐บ๐ผ๐ฟ๐ฒ:
Nuro builds the future of autonomous robotics with a new Google transfer feature โ https://goo.gle/4fd2Gp1
Case Study: How Nuro uses Google Cloud infrastructure for AI โ https://goo.gle/3y0Qneo