Evident Machines
Introducing
FrontierEE Bench
The first benchmark to test AI models at a physical electronics lab bench.
Explore the results →Autonomous hardware R&D is the next frontier for AI, and it begins at the lab bench.
Advancements in electrical engineering determine how much computation we can run, how efficiently we deliver power, and what our machines can do. From semiconductors to energy infrastructure and robotics, better hardware unlocks progress across the economy. Accelerating that work is one of AI’s most valuable opportunities.
But training in simulation alone leaves models unexposed to the many failure modes they’ll encounter in the physical world. Measurements are noisy, components vary, and hardware behaves in ways the simulation never captured.
For frontier models to become stewards of economic progress in EE, they must be trained on the realities of a physical lab bench, where real hardware puts their scientific judgment to the test.
Evident Machines turns experimental judgment into training data for frontier AI.
We build real-world EE benchmarks and RL environments on physical lab benches.
In our San Francisco lab, models work through electrical engineering tasks that require physical experimentation. They direct experiments, control instruments, and interpret measurements, while human operators handle wiring and probe placement.
Our benchmarks expose where models’ scientific judgment breaks down on real hardware. We capture expert workflows and build training data and RL environments around those failures, so models can learn from experimental outcomes.
We’re making physical experimentation part of how AI learns, starting with the engineering that powers the economy, and AI itself.
Work with us
We’re looking for scientists, engineers, and researchers in electrical engineering to help build meaningful evaluations from work they know deeply. If you’re interested in shaping the benchmark or collaborating on a research task, , or contact founders@evidentmachines.com.
To contribute to our benchmark paper, see our call for contributions.