Uploaded April 2026 | Updated September 2026, 2 weeks ago
"Every vehicle is capable of driverless operation. That's clearly the steady state of where we're going."
Wayve started in a rented house in Cambridge with $1.5M, a car in the garage, and an aim to integrate end-to-end AI into driving. A decade later it's driven across 506 cities without a single HD map and is worth over $8.6 billion.
In this episode, CEO Alex Kendall joins Lukas Biewald to talk about how he built the AI driver Uber, Nvidia, Mercedes, and Nissan all backed, and why putting self-driving AI into 100 million cars a year is a far bigger bet than 10,000 robotaxis.
Waymo and Tesla both come up. He doesn't shy away.
Connect with us here:
Alex Kendall: linkedin.com/in/alexgkendall
Lukas Biewald: linkedin.com/in/lbiewald
Wayve: linkedin.com/company/wayve-technologies
Weights and Biases: linkedin.com/company/wandb
(00:00) Trailer
(01:01) Introduction
(01:42) The story of Wayve
(02:44) End-to-end deep learning for autonomous vehicles
(03:52) Driving zero-shot in 500+ cities
(04:22) Growing up on a farm in New Zealand
(05:11) Why the company is called Wayve
(06:10) Building the first end-to-end prototype
(07:28) First lane following with 10 interventions
(09:09) Scaling beyond on-policy learning
(10:37) Why London was a great training ground
(11:15) The autonomous vehicle market then and now
(12:46) Mass market vehicles versus expensive retrofits
(14:15) Algorithm improvements versus logistics
(16:15) Plugging perception into an LLM
(17:07) Adding language to the driving model
(18:04) Learning social cues like flashing lights
(18:42) Personalization and driving styles
(19:23) Waymo's impact on San Francisco
(20:30) Human-like driving behavior
(20:49) Why Wayve doesn't build its own cars
(22:53) Stack ranking business opportunities
(24:00) Target geographies for launch
(25:08) The AI 500 road show results
(26:44) Is this superhuman performance?
(27:56) Scenarios that still challenge the model
(29:20) The trolley problem in self-driving
(30:53) Minimal risk maneuvers explained
(31:31) Training data and crash scenarios
(33:08) Handling regional driving norms
(34:28) World models versus policy learning
(35:18) Transitioning from CTO to CEO
(37:57) Advice to your younger self
(40:56) Algorithms versus infrastructure
(42:02) Evaluation and simulation challenges
(43:02) Taking Bill Gates for fish and chips
(44:06) Parties at the first prototype house
(45:36) Wrap-up
"Every vehicle is capable of driverless operation. That's clearly the steady state of where we're going."
Wayve started in a rented house in Cambridge with $1.5M, a car in the garage, and an aim to integrate end-to-end AI into driving. A decade later it's driven across 506 cities without a single HD map and is worth over $8.6 billion.
In this episode, CEO Alex Kendall joins Lukas Biewald to talk about how he built the AI driver Uber, Nvidia, Mercedes, and Nissan all backed, and why putting self-driving AI into 100 million cars a year is a far bigger bet than 10,000 robotaxis.
Waymo and Tesla both come up. He doesn't shy away.
Connect with us here:
Alex Kendall: linkedin.com/in/alexgkendall
Lukas Biewald: linkedin.com/in/lbiewald
Wayve: linkedin.com/company/wayve-technologies
Weights and Biases: linkedin.com/company/wandb
(00:00) Trailer
(01:01) Introduction
(01:42) The story of Wayve
(02:44) End-to-end deep learning for autonomous vehicles
(03:52) Driving zero-shot in 500+ cities
(04:22) Growing up on a farm in New Zealand
(05:11) Why the company is called Wayve
(06:10) Building the first end-to-end prototype
(07:28) First lane following with 10 interventions
(09:09) Scaling beyond on-policy learning
(10:37) Why London was a great training ground
(11:15) The autonomous vehicle market then and now
(12:46) Mass market vehicles versus expensive retrofits
(14:15) Algorithm improvements versus logistics
(16:15) Plugging perception into an LLM
(17:07) Adding language to the driving model
(18:04) Learning social cues like flashing lights
(18:42) Personalization and driving styles
(19:23) Waymo's impact on San Francisco
(20:30) Human-like driving behavior
(20:49) Why Wayve doesn't build its own cars
(22:53) Stack ranking business opportunities
(24:00) Target geographies for launch
(25:08) The AI 500 road show results
(26:44) Is this superhuman performance?
(27:56) Scenarios that still challenge the model
(29:20) The trolley problem in self-driving
(30:53) Minimal risk maneuvers explained
(31:31) Training data and crash scenarios
(33:08) Handling regional driving norms
(34:28) World models versus policy learning
(35:18) Transitioning from CTO to CEO
(37:57) Advice to your younger self
(40:56) Algorithms versus infrastructure
(42:02) Evaluation and simulation challenges
(43:02) Taking Bill Gates for fish and chips
(44:06) Parties at the first prototype house
(45:36) Wrap-up
