About
Making agents dependable enough for everyday life
EvoQuake CO. was founded by Michael Luo to close the gap between what agents can demo and what they can be trusted to do every day.
AI has crossed a threshold: models can now reason, use tools, and act. But capability isn't the same as dependability. Turning raw model ability into agents people rely on takes infrastructure — runtimes, guardrails, evaluation, and interfaces built for autonomy from day one.
That's the work. We build the layer between frontier models and the moments where they actually help someone get through their day.
Founder
Michael Luo
Michael founded EvoQuake after more than 15 years building the serving, data, and machine-learning infrastructure behind some of the world's largest recommendation systems. He's led engineering teams at Meta, Google, and Microsoft — the through-line being large, complex ML systems that have to stay fast, correct, and affordable at scale.
Most recently, as a Staff Software Engineer Manager on Meta's Recommendation System ML Foundation, he led the push toward AI-first development — automating multi-agent workflows to continuously improve the codebase and building agent tooling that removed operational friction for hundreds of ML engineers. At Google he spent nearly a decade on YouTube's discovery infrastructure, leading a 2+ year redesign of the core recommendation data model and pioneering storage systems that cut cost across hundreds of petabytes.
That experience — where AI capability meets the unglamorous systems work that makes it dependable — is exactly the gap EvoQuake exists to close.
- Previously
- Meta · Google / YouTube · Microsoft
- Focus
- ML, data & recommendation infra at scale
- Education
- MBA, Old Dominion · B.S., Tsinghua University
- Based in
- Bay Area, California
What we value
How we work
Reliability over spectacle
A flashy demo is easy. Software people depend on is hard. We optimize for the second one.
Empower, don't replace
Agents should expand what a person can do — not quietly take the wheel. People stay in control.
Build in the open
Composable, inspectable infrastructure beats black boxes. Trust comes from being able to look inside.