LCLS AI/ML, Algorithms
Computational Imaging – AI4Science
- AtomicSPI
- cryoAI
- aNiMAte
- ADP-3D
- cryoFIRE
- cryoDRGN-AI
- MorphOT
- X-RAI
- Physics-Guided Dual Implicit Neural Representations for Source Separation
- DynaBoost
- Nanobragg-torch (in development): differentiable simulator for protein crystallography.
- Eryx-torch (in development): differentiable simulator for diffuse scattering from protein crystals.
- crystalBoltz (in development): inference-time scaling of diffusion models for structural ensemble refinement during crystallography experiments
- AlphaEnsemble (in development): foundation model for multimodal biomolecular imaging
- Abismal
Automation
- BayFAI
- AutoMFX
- CoyoteProtector (in development): computer vision for detector protection during small molecule SFX experiments.
- An Agentic Artificially Intelligent X-ray Scientist
- Implicit neural representations for experimental steering of advanced experiments
- Bayesian experimental design and parameter estimation for ultrafast spin dynamics
Data Reduction/Feature Extraction
- ARAMS: Accelerated Rank Adaptive Matrix Sketching algorithm to facilitate the analysis of large-scale datasets
- PyPCA: Principal Component Analysis on imaging datasets of arbitrary size opens the door to rapid featurization with potential for experiment steering.
- PumpProbeDiagnostics (LUTE workflow; in development): supervised classification of water scattering signal for temperature-jump diagnostics.
- TimeTool-ML (smalldata workflow; in development): confidence assessment and signal boosting for LCLS-II timetool analysis.
- PeakNet: End-to-end deep learning pipeline for real-time Bragg peak segmentation: from training to large-scale deployment
- SpeckleNN: A unified embedding for real-time speckle pattern classification in X-ray single particle imaging with limited labeled examples
- SLAC Neural Network Library: SpeckleNN implemented in FPGA using SLAC Neural Network Library
- cuSZ Compression (led by Franck Cappello at Argonne)
Scalable Software
- CuPyNumeric: a distributed and accelerated drop-in replacement for the numPy API (Alex Aiken/Seshu Yamajala with NVIDIA)
- Legate: (Alex Aiken with NVIDIA) a framework and runtime with an ecosystem of libraries that democratize distributed accelerated computing for everyone.
Data Services
- LCLStream: The LCLStream Ecosystem for Multi-Institutional Dataset Exploration
- Data Engine: End-to-end deep learning pipeline from training to large-scale deployment.
- elog-copilot: This natural language interface enables researchers to intuitively search LCLS metadata, enrich LCLS metadata, analyze instrument performance trends, and uncover cross-experiment patterns, transforming petabytes of operational data into actionable knowledge through integrated LLM technology and FAIR data infrastructure.
