Actionable Information from Sensor to Data Center

Actionable Information from Sensor to Data Center

The project aims to develop an integrated AI pipeline that links edge-based inference systems with high-performance computing for real-time adaptive analysis of massive data streams from next-generation light sources, enabling automated extraction of actionable scientific insights that can dynamically guide experimental facilities.

Key Objectives

  • Develop a scalable AI-powered data pipeline to process the unprecedented volume and velocity of data from LCLS-II and APS-U light sources.
  • Implement edge computing infrastructure with AI accelerators for real-time inference on streaming experimental data.
  • Create a bidirectional link between edge devices and HPC systems that enables continuous model training and updating during experiments.
  • Design adaptive ML/AI algorithms that automatically tune to changing experimental conditions without manual parameter adjustment.
  • Enable real-time extraction of actionable information to drive dynamic facility responses and experimental steering.

Project Details

The AISDC project addresses the fundamental challenge of extracting scientific insights from the massive data streams generated by next-generation light sources. As LCLS-II and APS-U produce orders of magnitude more data than their predecessors, traditional analysis methods that require manual parameter tuning become unsustainable. AISDC creates a novel information pipeline that combines edge computing for immediate inference with high-performance computing for sophisticated model training. By implementing a closed-loop system where AI models continuously learn and adapt to new experimental conditions, AISDC eliminates bottlenecks in data analysis workflows. The system deploys specialized hardware accelerators (CPUs/GPUs, FPGAs, AI processors) at the edge to process streaming data in real-time, while simultaneously leveraging HPC resources for complex model training and refinement. This integrated approach enables experiments to dynamically respond to emerging data patterns, automatically adjust parameters, and extract physically meaningful information without human intervention. AISDC will transform light source experiments from data-limited to insight-driven operations, maximizing scientific returns from these advanced facilities across disciplines from materials science to structural biology.

Partners & Collaborators

Jana B. Thayer (SLAC)

Ryan Coffee (SLAC)

Frederic Poitevin (SLAC)

Deeban Ramalingam (SLAC)

Kevin Dalton (SLAC)

Cong Wang (SLAC)

Stephen Anthony Cisneros (SLAC)

Abhilasha Dave (SLAC)

Dionisio Doering (SLAC)

James Russell (SLAC)

Venkata Devesh Reddy Seethi (ANL)

Rajkumar Kettimuthu (ANL)

Weijian Zheng (ANL)

Dishant Beniwal (ANL)

Hemant Sharma (ANL)

Antonino Miceli (ANL)

Peter Kenesei (ANL)

Sebastian Strempfer (ANL)

Xiaolong Ma (ANL)

Viktoriya Yarema (ANL)

Gabriel Ponon (ANL)

Publications

[1] C. Wang, V. Mariani, F. Poitevin, M. Avaylon, and J. Thayer, “End-to-end deep learning pipeline for real-time bragg peak segmentation: From training to large-scale deployment”, Front. High Perform. Comput., vol. 3, 2025, Art. no. 1536471, doi: 10.3389/fhpcp.2025.1536471.

[2] W. Zheng, J.-S. Park, P. Kenesei, A. Ali, Z. Liu, I. Foster, N. Schwarz, R. Kettimuthu, A. Miceli, and H. Sharma, “Rapid detection of rare events from in situ X-ray diffraction data using machine learning”, J. Appl. Cryst., vol. 57, no. 4, 2024, doi: 10.1107/S160057672400517X.

[3] R. Herbst, R. Coffee, N. Fronk, K. Kim, K. Kim, L. Ruckman, and J. J. Russell, “Implementation of a framework for deploying AI inference engines in FPGAs”, in Accelerating Science and Engineering Discoveries through Integrated Research Infrastructure for Experiment, Big Data, Modeling and Simulation, K. Doug, G. Al, S. Pophale, H. Liu, S. Parete-Koon, Eds. Cham, Switzerland: Springer Nature, 2022, pp. 120–134.

[4] “SLAC NeuralNet Library Source”, GitHub, 2023. [Online]. Available: https://github.com/slaclab/snl/releases/tag/v0.2.0.

[5] N. Layad, Z. Liu, and R. Coffee, “Open source implementation of the CookieNetAE model”, GitHub. [Online]. Available: https://github.com/AISDC/CookieNetAE.

[6] C. Wang, P.-N. Li, J. Thayer, and C. H. Yoon, “PeakNet: An Autonomous Bragg Peak Finder with Deep Neural Networks”, arXiv:2303.15301 [physics.comp-ph], Mar. 2023. [Online]. Available: http://arxiv.org/abs/2303.15301.

[7] C. Wang, E. Florin, H.-Y. Chang, J. Thayer, and C. H. Yoon, “SpeckleNN: a unified embedding for real-time speckle pattern classification in X-ray single-particle imaging with limited labeled examples”, IUCrJ, vol. 10, pp. 568–578, 2023, doi: 10.1107/S2052252523006115

[8] C. Benmore, T. Bicer, M. K. Y. Chan, Z. Di, D. Gürsoy, I. Hwang, N. Kuklev, D. Lin, Z. Liu, I. Lobach, Z. Qiao, L. Rebuffi, H. Sharma, X. Shi, C. Sun, Y. Yao, T. Zhou, A. Sandy, A. Miceli, Y. Sun, N. Schwarz, and M. J. Cherukara, “Advancing AI/ML at the advanced photon source”, Synchrotron Radiat. News, vol. 35, pp. 28–35, 2022.

[9] H. Sharma, J.-S. Park, P. Kenesei, J. Almer, Z. Liu, and A. Miceli, “Speeding up diffraction analysis using machine learning”, in Acta Crystallogr. A: Found. Adv., vol. 78, 2022, pp. A141–A141.

[10] Z. Liu, H. Sharma, J.-S. Park, P. Kenesei, A. Miceli, J. Almer, R. Kettimuthu, and I. Foster, “BraggNN: fast X-ray Bragg peak analysis using deep learning”, IUCrJ, vol. 9, no. 1, 2022, doi: 10.1107/S2052252521011258

[11] A. Ali, H. Sharma, R. Kettimuthu, P. Kenesei, D. Trujillo, A. Miceli, I. Foster, R. Coffee, J. Thayer, and Z. Liu, “fairDMS: Rapid model training by data and model reuse”, in 2022 IEEE Int. Conf. Cluster Comput. (CLUSTER), Los Alamitos, CA, USA, 2022, pp. 394–405, doi: 10.1109/CLUSTER51413.2022.00050.

[12] M. Levental, A. Khan, R. Chard, K. Yoshi, K. Chard, and I. Foster, “OpenHLS: High-level synthesis for low-latency deep neural networks for experimental science”, arXiv:2302.06751 [cs.AR], Feb. 2023. [Online]. Available: https://doi.org/10.48550/ARXIV.2302.06751.

[13] Z. Liu, A. Ali, P. Kenesei, A. Miceli, H. Sharma, N. Schwarz, D. Trujillo, H. Yoo, R. Coffee, N. Layad, J. Thayer, R. Herbst, C. Yoon, and I. Foster, “Bridging data center AI systems with edge computing for actionable information retrieval”, in 3rd Annu. Workshop Extreme-scale Experiment-in-the-Loop Comput. (XLOOP), 2021, pp. 15-23.