Safety-Critical MVP Framework
by Shweta Shrivastava • Senior Director of Product Management at Waymo
Shweta is currently leading product management at Waymo, focusing on autonomous driving behavior, simulation tools, and ride-hailing commercialization. Previously, she was CPO at Nauto and held leadership roles at AWS and Cisco.
🎙️ Episode Context
Shweta discusses the unique challenges of product management for autonomous vehicles, contrasting 'move fast and break things' with safety-critical systems. She details how Waymo builds trust through 'natural' driving behaviors and shares leadership lessons on communication efficiency and career growth from her time at Amazon and Waymo.
Problem It Solves
How to apply Agile/MVP concepts to hardware or AI products where failure poses physical risks (unlike pure SaaS).
Framework Overview
Redefining the 'Minimum' in MVP for deep tech. Instead of launching with bugs to fix later, the product must meet a 'super-human' safety benchmark before public deployment, shifting the focus from speed-to-market to rigorous simulation and validation.
🧠 Framework Structure
Benchmark against human performance: ...
Zero compromise on safety: The 'Minim...
Simulation before Real World: Extensi...
Operational Metrics focus: Track 'sto...
When to Use
When building products in healthcare, autonomous systems, or heavy industry where errors have high liability or physical consequences.
Common Mistakes
Treating safety features as 'v2' items or assuming users will forgive critical failures as part of a 'beta' test.
Real World Example
Waymo does not rely on users to 'take over' in emergencies (L4 autonomy); the system is validated to handle all situations safely before launch, unlike L2 driver-assist systems.
The concept of MVP, which is so widely popular in the SaaS product management world... has a whole new meaning here at Waymo... can't really cut corners on safety.
— Shweta Shrivastava