Google Cloud has released a new video and case study on how we are using Gemini to accelerate materials discovery at the National Laboratory of the Rockies (NLR). The work builds on our Autonomous Multi-Modal Characterization and Processing (AM³) platform, which brings agentic AI to the instrument.
The bottleneck we kept running into was cognitive, not mechanical: our instruments could act faster than a researcher could decide what to do next. Researchers can now state a goal in plain language, and the system proposes a plan, shows its reasoning, and waits for approval before anything moves. Its decisions are grounded in instrument documentation and past experiments, while hardware control stays within a defined set of approved commands.
In practice, calibration has fallen from more than 90 minutes per condition to about 13, and focusing tasks that once took up to 55 manual adjustments now take two. Just as importantly, computed metrics are replacing subjective visual judgment, giving us measurements we can reproduce. The researcher remains at the center of every consequential decision, with more attention available for the science itself.
Thank you to our team and to our colleagues at Google Public Sector. We look forward to building on this work together.