Infrastructure to Enable Shared Data & Computing

Trustworthy, privacy‑preserved federated learning for science

The models travel between institutions. The sensitive data never does.

Session held · Tuesday, June 2, 2026

Eight silos, one aggregator. Only model updates cross the wire.

When: Tuesday, June 2, 202614:00
Where: Annapolis & ColumbiaBaltimore, Maryland
Format: Birds-of-a-Feather8 lightning talks + discussion
Track: Shared InfrastructureTrillion Parameter Consortium

The session

Bringing together the people building federated learning for scientific workloads

Federated learning offers a promising approach for enabling collaborative scientific discovery while preserving the privacy of sensitive data across institutions. This BOF brought together researchers and practitioners to discuss trustworthy, privacy-preserving federated learning frameworks tailored for scientific workloads.

The discussion focused on challenges such as secure model aggregation, data confidentiality, system scalability, and integration with distributed research infrastructures. Our objectives were to identify common requirements, share emerging techniques, and foster collaborations within the TPC community — work that continues past the session itself.

The session speaks to anyone in TPC working on distributed computing, secure data sharing, and scalable AI methods that support cross-institutional scientific research.

Secure model aggregation

Combining updates from many institutions without any single party — including the aggregator — learning more than it should.

Data confidentiality

Differential privacy, ownership protection, and the guarantees a data steward actually needs before agreeing to participate.

System scalability

What breaks when a federation spans supercomputers, and what it takes to train a foundation model across facilities.

Integration with research infrastructure

Fitting federated workflows into the authentication, allocation, and data-governance realities of national labs and hospitals.

Organized by

Session leads

Lead organizer Olivera Kotevska Oak Ridge National Laboratory
Co-organizer Kibaek Kim Argonne National Laboratory
Co-organizer Ravi Madduri Argonne National Laboratory

Lightning talks

Running order

Eight talks, spanning brain-health networks, cross-facility training on supercomputers, and production federated learning in industry. Slides are posted where the speaker has shared them.
No.	Talk	Speaker	Affiliation	Slides
01	NeuroFL: OBI’s Intelligence Network for Brain Health	Francis Jeanson	Ontario Brain Institute	PDF
02	The Next Frontier: Federated AI with Flower	William Lindskog-Munzing	Flower Labs	—
03	Socio-Technical Infrastructure: Operationalizing FL Systems	Mohammed Manzari	Deloitte	—
04	Are You Ready for Production Federated Learning?	Holger Roth	NVIDIA	PDF
05	Federated LLM Training Across NNSA Labs	Max Carlson	Sandia National Laboratories	PDF
06	Scalable Cross-Facility Federated Learning for Scientific Foundation Models on Multiple Supercomputers	Yijiang Li	Argonne National Laboratory	—
07	OmniFed: Towards Configurable Cross-Silo Federated Learning	Sahil Tyagi	Oak Ridge National Laboratory	PDF
08	Differentially Private Federated Averaging with James–Stein Estimator	Minseok Ryu	Arizona State University	—

Speakers: to add your deck here, send it to kotevskao@ornl.gov.

Get involved

The conversation didn’t end in Baltimore.

We are continuing to build a community around trustworthy, privacy-preserving federated learning for science — shared requirements, common benchmarks, and cross-institutional pilots. If you work in this space, or want to bring your facility’s data into a federation without giving it up, get in touch.

Email the organizers kotevskao@ornl.gov