About Atomic Metrix
A project built by cyclists who wanted training advice to adapt to real humans, not fixed formulas.
Atomic Metrix started in September 2024 from repeated conversations among endurance riders who cared deeply about scientific training, but kept running into the same limit: conventional training metrics are static, while recovery, adaptation, and progress are not.
Team
ZWZhimin Wang
Zhimin holds a master's degree in Bioengineering from the University of Illinois Urbana-Champaign. He is passionate about endurance sports, familiar with scientific training, and has experience in software infrastructure and systems engineering.
Project story
We are building a training system that treats adaptation as personal, time-varying, and context-dependent.
We started Atomic Metrix after years of discussing scientific training from the perspective of real riders. Metrics such as FTP, TSS, and recovery heuristics are useful, but they are still fixed abstractions. Two athletes can complete the same workout and leave with very different fatigue, readiness, and improvement potential.
The emergence of modern foundation models gave us a new way to design training software. Instead of applying a rigid template, we can combine ride data, context, goals, and prior training history to build an agent that produces more individualized plans and more useful guidance.
Atomic Metrix is the project we built from that belief. We want to turn scattered consumer training data into a health agent that can reason about load, recovery, and progression with enough nuance to actually help athletes make better decisions.
Principle 1
Human adaptation is personal
We do not assume two riders should respond identically to the same load.
Principle 2
Recovery matters as much as effort
Better guidance requires understanding whether a rider is ready to absorb the next block.
Principle 3
Agents should personalize, not generalize
We use an agent layer to make training guidance more specific, contextual, and actionable for each athlete.