Real-time ML Inside a $2M Electron Microscope for Nuclear Powered Data Centers | Theia Scientific

Chris and Kevin Field from Theia Scientific show how they built a Grafana-powered platform to run real-time machine learning inside a $2 million electron microscope — helping validate materials for fusion-powered data centers. Their system won the 2024 GrafanaCON Golden Grot award for most beautiful dashboard.

Fusion energy reactors produce neutrons that create nano-scale damage in structural materials. To prove those materials will survive for decades, researchers irradiate samples with ion beams at facilities like the Michigan Ion Beam Laboratory and watch damage appear in real time under a transmission electron microscope. The old workflow: graduate students manually circling every defect dot in ImageJ — software unchanged since Windows 95.

Theia Scientific replaced that with the Theiascope: an on-premise GPU device running Grafana and Jupyter together, air-gapped from the internet, physically attached to the microscope. ML models trained on real and synthetic microscopy images run inference live during the experiment, populating Grafana dashboards with uniformity charts, per-layer confidence scores, and batch quality metrics as each image is captured.

The architectural key: a custom Grafana data source that passes Python code through Jupyter kernels, making Jupyter Notebooks function as a native Grafana data source. Scientists write Python directly in the panel query editor — the language they already know — and see results update live without learning SQL. Dashboard variables pass through to notebook context, enabling parameterized on-the-fly analysis during the experiment itself. The system is now deployed at Idaho National Lab and Argonne National Lab.

0:00 Introduction
0:44 Fusion Energy 101: Materials, Neutrons, and Data Centers
2:03 Ion Beams and Electron Microscopy for Fusion Research
4:11 The Problem: Manual Analysis with ImageJ
4:52 ML for Automated Defect Detection
6:46 Introducing the Theiascope: Grafana + Jupyter On-Device
7:22 Live Demo: Real-Time TRISO Particle Analysis
12:59 Running Jupyter Notebooks as Grafana Data Sources
15:11 Why Grafana for Scientific Instrumentation?
16:44 Grafana Is RAD: Rapid Application Development for Scientists
21:26 Conclusion and Collaborators

Links/resources:
Learn about Theia Scientific: https://www.theiasc.com
Check out GrafanaCON: https://www.grafana.com/grafanacon/?src=yt
Get started with the Grafana Cloud forever-free tier: https://grafana.com/g/cloud
Have a question? Ask Grot, your AI helper: https://grafana.com/grot/
Reach out in our community forums: https://gra.fan/communityyf

---
Thanks for watching!

👍 Was this video helpful? Like and subscribe to our channel for more videos.

Connect with Grafana Labs:
X: (https://www.twitter.com/grafana)
LinkedIn: (https://www.linkedin.com/company/grafana-labs/)
Facebook: (https://www.facebook.com/grafana)

#Grafana #Observability #NuclearScience #DataCenter #FusionResearch #Fusion #JupyterNotebook