A safety-critical physical AI service, such as a fully autonomous ride service
Build an AI agent that operates in the physical world at scale by combining a robust onboard agent, realistic simulation, systematic evaluation, and a deployment process designed around safety. The business depends on bridging the large gap between a working demonstration and a reliable product that can operate publicly without a human behind the wheel.
physical B2B Deep Tech / Infrastructure
From Y Combinator — Waymo Co-CEO Dmitri Dolgov: The Demo Is Only 1% Of The Work at 08:08
Problem: Provides useful physical-world autonomy where errors cannot be undone, latency matters, physical-world data is scarce, and the system must achieve very high safety and reliability before public deployment.
For: Members of the public who use an autonomous ride service, as well as future commercial customers for physical AI applications such as trucking and personally owned vehicles.
Examples
- Waymo autonomous driving: a team of about a dozen engineers achieved 100,000 autonomous miles and completed ten 100-mile routes without human intervention in about eighteen months, but took about fifteen years to reach the first 100 million autonomous miles and about seven months to drive the next 100 million.
Behind this: 15 build steps · 7 tools and how each is used · how to validate demand · 4 more real examples · 9 things the video never answers.
Other takes on Robotics operations and support services
- Robotics application company that solves one narrow physical-world business problem end to end
- Autonomous underwater robotics platform for ecosystem restoration, offshore industry, and maritime defense
- Open humanoid-robot platform for household and commercial workflows
- Humanoid robot automation for warehouse and other human workflows
- General-purpose robotic automation for real-world business workflows
- Industrial robotic inspection for critical infrastructure and hazardous facilities