EPSCI

Group Meeting�

Sep 15, 2026

1

2

LD2606 and AmSC Data Services - Data Movement

Sep 2 – Sep 15 Accomplishments

1. Publications:
• SC26 PDSW workshop. DAOS paer was rejected. Resubmit it as a 1-page WIP and pray for luck.
• IPDPS’27: June 1–5, 2027, Seattle. Abstract ddl: Oct-1 AOE; Paper ddl: Oct-8 AOE. GPU Variabbility. Last 20% effort.
• AMD MI250 / Intel Max 1550 GPU memory performance is not as consistent as NVIDIA A100.
• GPUs are suffering from power walls.
• LDRD wrapup. Some conference in 2027/2028. Discuss with collaborators soon.
2. AmSC Data Infra Project will have a second year.
3. FY25–FY26 LDRD - continuous real-time subsecond network monitoring, visualization and replay for distributed 100 Gbps TCP/IP env.
• eBPF-based. Works for any modern Linux machine (including VM) with sudo
• Pairwaise iperf3 test passed. Record.
4. FY27 – FY28 LDRDs, help Jon & Cameron.
5. NSF GM collaboration letter.

EPSCI Group Meeting Sep 15, 2026

EIC Computing and Software

3

Accomplishments

1. one …
2. two …
3. three …

Production

• Thomas has stepped back from convening the production WG due to other time commitments

Validation

• Datacards: Genesis mission datacards and simpler ePIC datacards are under construction as part of the ePIC data lifecycle management task force.
• Transitioning towards Rucio for sending Hydra pipeline images
• Set up ePIC virtual machine to run Hydra system
• MVP for data lifecycle under development
• Links data cards, snakemake workflows, and hydra into a coherent system for data management and reusability in ePIC
• Working on slides to present MVP at ePIC software and computing review at BNL

EPSCI Group Meeting Sep 15, 2026

JANA2/LD2613

4

LD2613

1. Experimenting more with agentic AI assisted development
2. Closing out remaining Q4 milestones
• Added MCP server!
• Shell runs locally; connects to Conductor via MCP server
• Integrations with code.jlab.org (integration testing/preprod/prod)
• Conductor VM automatically reloads from container repo
• Container automatically rebuilds on CI runner
• Next: CI runner launches workflow via Conductor VM (“preprod”)
3. Goal: Demo by end of week

�JANA2

1. Quiet time for JANA
2. ​

Raiqa

• Added detector_translation layer to jana2_common_extensions
• Working on extracting translators into a plugin

EPSCI Group Meeting Sep 15, 2026

DAQ: Streaming Readout + CODA

5

Hall B SRO Status

1. Had check in with Sergey B last Friday
2. Main 18 crates not ready
• CentOS upgrade of VME in progress �(maybe ready this week?)
• Recompile rol, test that DAQ works again
3. Counting house and hycal crates available for testing
4. ejfat-1 status?

Software Work

• GEMC discussion with Maurizio
• Emu pull request from Vardan
• SC live cross-site demonstration �(Cissie, Vardan, Jon, others?)
• Work with Cole Smith and calorimeters to begin SRO recon?

Upcoming Dates

• LDRD due Sept 21. Most comments about connecting better to physics and EIC
• Jon will give JLab epsci sro update for SRO XIV (EIC, end of month)
• Poster for SC26 accepted, upload final version by Sept 30

EPSCI Group Meeting Sep 15, 2026

Genesis Mission / American Science Cloud

6

Accomplishments

1. one …
2. two …
3. three …

BPSW

• Contribution to Knowledge Base merged
• Gearing up for the hackathon
• Should know what I am going to be helping with by the 25th

Raiqa - Mya AmSc

• MVP Core Functionality: Globus data transfer pipeline for MYA data is complete
• TODO left: Limit to JLab and BNL only through globus authentication
• Deployment & Security: Web page is live on the VM and passed cyber review; pending final firewall access to make it publicly accessible.
• Data Integration: MVP currently streams demo data. Work on Data Card and review form for the AmSci Data Catalog is in progress with a alignment meeting scheduled with Diana this Wednesday.

EPSCI Group Meeting Sep 15, 2026

Coupling Experiment to Accelerator Controls

7

TwinD

1. reBenchmarking ongoing
• Technical hiccups
2. Added pricing statistics including expected cost per success
3. Added a global comparison
4. Generating a “run list”

GlueX

• Implemented anomalous EPICS producer and anomalous GP agent into SMOCS. Can inject anomalies into EPICS messages to observe model behavior for CDC HV settings.
• Neural network for TAGM timewalk corrections re-implemented using pytorch, doing standard validation and verification before integrating into halld_recon
• Deployment on gluon200 in Hall-D counting house ongoing
• BCAL classifier was retrained after removing low feature-importance variables, performance was unchanged
• L3 Trigger work: Sergey has resumed FPGA implementation post-vacation, hardware issues are being addressed, and possible implementation of GNN for classification being discussed

MOLLER

• Set up gitlab repository to better manage issues, tracking progress, and general things that gitlab is useful for
• Harp scan data has been extracted from the logbooks, will be used to inform models of constraints on beam spot size
• Working with Kishan on formalized problem statement (optimization vs control)
• Addressing milestones and coming up with a revised set and work plan with Adam Carpenter

EPSCI Group Meeting Sep 15, 2026

Hydra

8

Accomplishments

1. one …
2. two …
3. three …

Developments

• PR backlog through
• Raiqa revamped frontend testing and deployment:
• Automated and standardized the end-to-end release workflow, reducing new deployment setup to a few configuration steps while minimizing missed steps, inconsistencies, and deployment failures.
• Added tested safeguards including preflight validation, health checks, immutable releases, atomic publishing, reversible test and production rollbacks, safe test resets, automatic failure recovery, service restart, and release cleanup.
• Integrated the frontend and backend repositories as submodules under epsci/hydra/web/stack and established versioned releases, with `v0.0.1` as the latest release.
• New version roll out

EPSCI Group Meeting Sep 15, 2026

Miscellaneous

9

Accomplishments

1. one …
2. two …
3. three …

EPSCI Group Meeting Sep 15, 2026