Anthony Kougkas, PhD
Associate Research Professor, Illinois TechGuest Research Faculty, ArgonneExecutive Director, Gnosis Research Center
I build systems, teams, and ideas that help people do ambitious work.
My field is the data behind scientific computing: how storage, I/O, and now AI get the right data, in the right form, at the moment it is needed.
The lineage behind IOWarp and CLIO
Browse all work- IRIS ICS'18. Files and objects behind one interface.
- Hermes HPDC'18, then NSF 2018–2023. Multi-tier I/O buffering.
- LABIOS HPDC'19 best paper, US patent 2023, NSF 2023–2027.
- ChronoLog MSST'20, then NSF 2021–2027. A time-ordered shared log.
- DeepIO Argonne, PNNL, and Livermore subcontracts from 2022. I/O meets AI.
- DTIO DOE 2023–2026, SSDBM'25. Labels become data tasks.
- IOWarp NSF 2024–2029. The platform under CLIO.
- CLIO Core, Kit, Agent, Relay, Coder. The context layer.
Current work
- IOWarp 2024–present
- The NSF-funded context-management program I proposed and architect across the Gnosis Research Center, The HDF Group, and the University of Utah. It runs from a laptop to Ares, DeltaAI, Polaris, and Aurora.
- CLIO 2025–present
- The architecture I created for IOWarp. Its shared contracts let Core, Kit, Agent, Relay, and Coder give scientific AI agents structured, queryable context. Luke Logan leads Core, Jaime Cernuda leads Agent and Relay, students build Kit, and I build CLIO Coder myself.
The software is public at iowarp.ai, with portals for CLIO Agent, CLIO Kit, CLIO Core, and CLIO Coder.
Recent notes
All notes- Nov 13, 2026
Reliable work from unreliable agents
A coding agent is a system built around one brilliant, persuasive, unreliable component. The argument of my November 9, 2026 lecture on agentic engineering: five ideas for building trustworthy work around that component, what they could do for science, and the problems still open.
- Nov 13, 2026
One destination, N tiers
Every tier supplies the application at once, and the application sees one destination. The thesis from a November 2026 talk to my group at the Gnosis Research Center, the six lines of I/O work behind it since 2016, and my bet that they are the foundation agents need for context.
- Nov 8, 2026
SC26 in Chicago
What the group brings to SC26 in Chicago, the pressure from GPUs and agent fleets that runs through all of it, and an invitation to Thursday's Birds of a Feather on HDF5 in the age of AI-driven science.
Papers behind the systems
-
Hermes
-
LABIOS
LABIOS: A Distributed Label-Based I/O System HPDC 2019 Karsten Schwan Best Paper
-
ChronoLog
-
DeepIO
-
DTIO
DTIO: Data Stack for AI-driven Workflows SSDBM 2025
At SC
- LabStor: A modular and extensible platform for developing high-performance, customized I/O stacks in userspace SC 2022
- Data Flow Lifecycles for Optimizing Workflow Coordination SC 2023
- MegaMmap: Blurring the Boundary Between Memory and Storage for Data-Intensive Workloads SC 2024
- DFTracer: An Analysis-Friendly Data Flow Tracer for AI-Driven Workflows SC 2024
- Trigger, Render, Reason: Multimodal LLM Agents for Science-Aware Scientific Simulation Monitoring SC 2026 Accepted
Citations are how systems researchers say thank you. Hermes has collected the most; the others are patiently waiting. Every paper page has one-click BibTeX.
In brief
My research is one line of questions, each opened by the answer before it. IRIS asked whether files and objects could sit behind one interface. Hermes asked who should decide where data lives once a machine has many tiers of memory and storage, and made placement the middleware's job. LABIOS asked what that middleware could do if each request said what it was for, and made the label carry intent. ChronoLog asked how to order the activity those requests leave behind, and chose physical time over consensus. DeepIO, my series of subcontracts with Argonne, Pacific Northwest, and Lawrence Livermore, asked what deep learning needs from the I/O stack, and funded the students who followed that question from training to inference. DTIO treated the label as a unit of work, a data task a runtime can compose and schedule. IOWarp asks the question all of them were circling: which data, metadata, tools, and history does a computation need right now? IOWarp carries Hermes placement, LABIOS intent, DTIO tasks, and ChronoLog provenance forward, and CLIO is the architecture that gives that context to scientific AI agents.
I co-founded the Gnosis Research Center with Xian-He Sun in 2023, where IOWarp and CLIO are built, and co-advise its PhD students; two of the first, Luke Logan and Jaime Cernuda, stayed on at Illinois Tech as research faculty.