Anthony Kougkas, photographed for the Gnosis Research Center member directory.

Anthony Kougkas, PhD

Associate Research Professor of Computer Science, Illinois Institute of TechnologyGuest Research Faculty, Argonne National LaboratoryCo-Founder and Executive Director, Gnosis Research Center

I build systems, teams, and ideas that help people do ambitious work.

I create and lead research programs across high-performance computing, storage and I/O, data systems, and AI. From Hermes and LABIOS to IOWarp and CLIO, I turn questions into proposals, architectures, teams, partnerships, papers, and working software.

Explore the work Read the journal

[ io = input / output ]

An input and output trace passing through a context field A restrained system diagram connecting data, tools, people, and evidence through a central context node. INPUT DATA CONTEXT EVIDENCE OUTPUT people tools
The useful work happens between input and output: context, judgment, and evidence.

Current research program / IOWarp + CLIO

Context infrastructure for AI at scale

IOWarp and CLIO carry HPC data-systems discipline into context infrastructure for scientific AI.

The IOWarp platform connecting scientific applications and AI workflows to data movement, transformation, and storage services.
IOWarp coordinates data movement, transformation, metadata, and storage across scientific workflows. CLIO turns those capabilities into a context architecture for people and agents.

I lead the research direction and architecture across IOWarp and CLIO, coordinate the work among people and institutions, and contribute across the software ecosystem.

IOWarp
The collaborative research program connecting the data-system lineage to context management for scientific AI.
CLIO
The context architecture spanning storage, transfer, discovery, scientific tools, agents, provenance, and human control.
Read the IOWarp story See the CLIO system

Research fabric / 2018–2026

From moving bytes to managing context

HPC, storage, I/O, and data management remain the foundation. Each project addresses a different systems problem; together they lead toward reliable context for AI.

Anthony Kougkas's research fabric, 2017 to 2027 Collaborative research programs run from IRIS and Hermes through LABIOS, ChronoLog, DLIO, DTIO, IOWarp, and CLIO. Public studio explorations run from WarpGPT through Warpio and Clio Coder. Dotted links mark ideas carried between the two. IRIS IRIS: ICS'18. Files and objects behind one interface. Hermes Hermes: HPDC'18, then NSF 2018–2023. Multi-tier I/O buffering. LABIOS LABIOS: HPDC'19 best paper, US patent 2023, NSF 2023–2027. ChronoLog ChronoLog: MSST'20, then NSF 2021–2027. A time-ordered shared log. DLIO DLIO: CCGrid'21 best paper. Now inside MLPerf Storage. DTIO DTIO: DOE 2023–2026, SSDBM'25. Labels become data tasks. IOWarp IOWarp: NSF 2024–2029. The platform under CLIO. CLIO CLIO: Core, Kit, Agent, Relay, Coder. The context layer. WarpGPT WarpGPT: The natural-language interface I wrote into the IOWarp proposal. Warpio Warpio: Three tries in one month. Scientific extensions for Claude Code and Gemini CLI, then Warpio CLI, a Gemini CLI fork with personas and context handover. Archived. Clio Coder Clio Coder: A public testbed for typed tools, bounded workers, sealed receipts, and local models.
Collaborative programs form the main research line. Studio experiments below test ideas at smaller scale and feed them back into IOWarp and CLIO.

Now / Aug 2026

Where my attention is

I am working on context infrastructure for AI that has to operate across scientific data, storage hierarchies, clusters, tools, and human decisions. IOWarp and CLIO are the larger research program. Clio Coder is one hands-on testbed for the contracts, provenance, and control that program needs.

Open the current snapshot

Research programs and systems

Different projects, connected questions

Five systems show how the research moved from storage hierarchy and I/O intent through time, data tasks, and AI workload behavior.

complete 2018–2024

Hermes

The multi-tier I/O buffering system I created during my PhD, carried through five years of NSF funding into maintained middleware at The HDF Group and forward into IOWarp.

paper Hermes at HPDC 2018
evolving 2017–present

LABIOS

What happens when an I/O request carries its own intent. My dissertation idea, a best-paper award at HPDC, a US patent, an NSF award, and an evolving label-based runtime for data movement and transformation.

paper LABIOS at HPDC 2019
active 2020–present

ChronoLog

A distributed shared log that orders activity data by physical time instead of consensus. My MSST 2020 design, funded by NSF for six years, and now at version 3 with SQL, pub/sub, and an MCP server.

paper ChronoLog at MSST 2020
complete 2023–2026

DTIO

A task-based I/O runtime built with Argonne under a DOE ASCR award. Labels became composable data tasks, online format translation with caching beat offline conversion by 49.6%, and the mechanisms were reimplemented inside CLIO Core.

paper DTIO at SSDBM 2025
active 2020–present

DeepIO

I/O for scientific AI, measured before optimized. DLIO, the benchmark my team built with Argonne, won a best-paper award and became the workload generator inside MLPerf Storage. The line continues through Viper, WisIO, and KV-cache scheduling for LLM inference.

paper DLIO at CCGrid 2021

All work stories

Personal studio

Where I build at close range

Tools, prototypes, and writing developed at close range. These explorations test ideas that can later strengthen the larger research programs.

Coding systems

Clio Coder

Local-first coding infrastructure for scientific software, built around typed tools, bounded workers, and durable receipts.

Research practice

Journal

Working notes on systems research, software, leadership, teaching, and the ideas that are still changing.

Tools and prototypes

Project atlas

Public software, small tools, and experiments alongside the collaborative research portfolio.

Recent journal

Notes from the workbench

All entries
essay 5 min read

Context is part of the system

A capable model still fails when the system hands it the wrong slice of reality. CLIO's answer is to treat context the way a database treats state, and to seal a receipt for every action.

essay 6 min read

When is research software shipped?

A working definition of "shipped" for research software, built around a named state someone else can run, evidence tied to that state, explicit limits, and a continuity signal, with Hermes and DLIO as the worked examples.

update 2 min read

DLIO joins MLPerf Storage

MLPerf Storage now uses DLIO, our deep-learning I/O benchmark, to generate its workloads. A research tool found a second life.

Anthony Kougkas with GRC students in front of the SC24 sign in Atlanta.
With the GRC team at SC24 in Atlanta.

The people

Ambitious systems need ambitious teams.

I build research programs with students, research staff, co-authors, institutional partners, and national laboratory collaborators. Leadership means giving people consequential ownership while keeping the whole architecture moving in one direction.

Meet the people behind the work