Towards an Integrated Research Infrastructure for Laser Plasma Acceleration Experiments

To support more efficient fusion ignition, we are developing a

digital twin, driven by AI/ML capable of providing real-time opti-

mization guidance, learning from both experimental and simulation

data. We have developed a dashboard hosted on the container-based

platform SPIN that provides real-time guidance for running optimal

laser-plasma experiments. Within this dashboard, we have integrated

methods to automate running GPU simulations and training ML

models using both simulation and experimental data using the LBNL

NERSC (National Energy Research Scientific Computing) Superfacil-

ity API (Application Programming Interface). We use a databse to

store experimental and simulation data and the ML model. From this

database, we visualize the ML model predictions with both simulation

and experimental data.