# Izzet Yildirim

> PhD candidate at Illinois Tech building the data-movement and I/O infrastructure that keeps GPUs fed, for AI training and LLM agents.

Izzet Yildirim is a PhD candidate in computer science at Illinois Institute of Technology (Gnosis Research Center, advisors Anthony Kougkas and Xian-He Sun), expected to graduate in May 2027. He worked with Lawrence Livermore National Laboratory's CASC from 2022 to 2026 and interned at Argonne National Laboratory in 2024. Before his PhD he spent more than ten years in industry, eight of them at VeriPark, where he rose from developer to leading a 20+ engineer division that shipped mobile banking to 3M+ users.

He is looking for AI-infrastructure and research-scientist roles starting Summer 2027, and is open to postdoctoral and fellowship positions.

## Pages

- [Research](https://iyildirim.com/research): all papers, workshop papers, and posters
- [CV](https://iyildirim.com/cv): experience, education, teaching, service, and skills
- [Notes](https://iyildirim.com/notes): short write-ups on agents, storage, and research
- [Contact](https://iyildirim.com/contact): email iyildirim@iyildirim.com

## MCP server

- https://iyildirim.com/api/mcp: read-only Streamable HTTP MCP server with tools for publications, paper full text, CV, and profile

## Publications

- [GDSight: Per-Operation Cross-Layer Attribution for GPUDirect Storage](https://iyildirim.com/research/gdsight). Izzet Yildirim, Xian-He Sun, Anthony Kougkas. 11th International Parallel Data Systems Workshop (PDSW), 2026. Per-operation, cross-layer attribution for GPUDirect Storage: where a single GPU read spends its time, from the application down to the device.
- [ORION: Observability Framework for Hybrid Workflow-Centric I/O Monitoring in HPC Systems](https://iyildirim.com/research/orion). Hariharan Devarajan, Ray A. O. Sinurat, Izzet Yildirim, Olga Kogiou, Aniruddha Marathe, Nikoli Dryden. 8th Workshop on Programming and Performance Visualization Tools (ProTools), 2026. An observability framework for hybrid, workflow-centric I/O monitoring on HPC systems. I co-authored it.
- [StageOnce: Speculative Cross-Job Data Staging for LLM Agent Fleets](https://iyildirim.com/research/stageonce). Izzet Yildirim, Xian-He Sun, Anthony Kougkas. Research Posters, International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2026. Speculative cross-job data staging for fleets of LLM agents.
- [ToolFuse: Provenance-Aware In-Flight Coalescing of Redundant Agent Tool Calls](https://iyildirim.com/research/toolfuse). Izzet Yildirim, Xian-He Sun, Anthony Kougkas. Research Posters, International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2026. Provenance-aware, in-flight coalescing of redundant agent tool calls.
- [AgentStage: Exploiting LLM Thinking for Data Staging in Scientific Agents](https://iyildirim.com/research/agentstage). Izzet Yildirim, Xian-He Sun, Anthony Kougkas. 22nd IEEE International Conference on e-Science (eScience), 2026. Uses an agent’s thinking time to stage data from cold storage into a tiered node-local cache (memory, then NVMe) before the agent asks for it.
- [Correct Is Not I/O Efficient: Characterizing I/O Behavior of LLM Agents on Scientific Data](https://iyildirim.com/research/correct-not-io-efficient). Izzet Yildirim, Jie Ye, Jaime Cernuda Garcia, Xian-He Sun, Anthony Kougkas. Greater Chicago Area Systems Research Workshop (GCASR), 2026. Characterizes the I/O behavior of LLM agents working on scientific data: an agent can reach a correct answer while doing far more I/O than the task needs.
- [Optimizing I/O for an Exascale Implicit Kinetic Plasma Simulation using the Rabbit Storage System](https://iyildirim.com/research/rabbit-io). Ian Lumsden, Hariharan Devarajan, Izzet Yildirim, Stefano Markidis, Andong Hu, Ivy Peng, Luca Pennati, Dewi Yokelson, Stephanie Brink, Olga Pearce, Tom Scogland, Bronis R. de Supinski, Gian Luca Delzanno, Anthony Kougkas, Xian-He Sun, Michela Taufer. 5th Workshop on Re-envisioning Extreme-Scale I/O for Emerging Hybrid HPC Workloads (REX-IO), 2025. I/O optimization for an exascale implicit kinetic plasma simulation on LLNL’s Rabbit near-node NVMe storage.
- [WisIO: Automated I/O Bottleneck Detection with Multi-Perspective Views for HPC Workflows](https://iyildirim.com/research/wisio). Izzet Yildirim, Hariharan Devarajan, Anthony Kougkas, Xian-He Sun, Kathryn Mohror. 39th ACM International Conference on Supercomputing (ICS), 2025. Automated I/O bottleneck detection for HPC workflows, using correlated views of the same trace from several perspectives.
- [Recorder: Comprehensive Parallel I/O Tracing and Analysis](https://iyildirim.com/research/recorder). Chen Wang, Izzet Yildirim, Hariharan Devarajan, Kathryn Mohror, Marc Snir. The Journal of Supercomputing, under review, 2025. Comprehensive parallel I/O tracing and analysis across HPC I/O libraries.
- [DFTracer: An Analysis-Friendly Data Flow Tracer for AI-Driven Workflows](https://iyildirim.com/research/dftracer). Hariharan Devarajan, Loïc Pottier, Kaushik Velusamy, Huihuo Zheng, Izzet Yildirim, Olga Kogiou, Weikuan Yu, Anthony Kougkas, Xian-He Sun, Jae-Seung Yeom, Kathryn Mohror. 36th International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2024. An analysis-friendly data-flow tracer for AI-driven workflows, developed at LLNL.
- [Exploring the Impacts of Multiple I/O Metrics in Identifying I/O Bottlenecks](https://iyildirim.com/research/io-metrics-sc23). Izzet Yildirim, Hariharan Devarajan, Anthony Kougkas, Xian-He Sun, Kathryn Mohror. Research Posters, International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2023. Shows that I/O time, bandwidth, and IOPS each expose different bottlenecks in HACC, Montage, and CM1.
- [IOMax: Maximizing Out-of-Core I/O Analysis Performance on HPC Systems](https://iyildirim.com/research/iomax). Izzet Yildirim, Hariharan Devarajan, Anthony Kougkas, Xian-He Sun, Kathryn Mohror. 8th International Parallel Data Systems Workshop (PDSW), 2023. Makes out-of-core analysis of very large I/O traces fast on HPC systems.
- [A Multifaceted Approach to Automated I/O Bottleneck Detection for HPC Workloads](https://iyildirim.com/research/bottleneck-detection-sc22). Izzet Yildirim, Hariharan Devarajan, Anthony Kougkas, Xian-He Sun, Kathryn Mohror. Research Posters, International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2022. DigIO, a tool that finds I/O bottlenecks automatically from a multifaceted view of a workload’s traces; the precursor to WisIO.
- [Performance Analysis of Containerized OrangeFS in HPC Environment](https://iyildirim.com/research/orangefs-sc21). Izzet Yildirim, Meng Tang, Anthony Kougkas, Xian-He Sun. Research Posters, International Conference for High Performance Computing, Networking, Storage, and Analysis (SC), 2021. Compares five container runtimes hosting the OrangeFS parallel file system; Singularity came closest to bare metal.
- Under review: The first I/O benchmark for scientific LLM agents. Details after publication.
- Under review: Online tuning of data movement for distributed training. Details after publication.
- Under review: Diagnosing and fixing deep learning I/O bottlenecks. Details after publication.

## Profiles

- [Scholar](https://scholar.google.com/citations?user=FmMwmlUAAAAJ&hl=en)
- [GitHub](https://github.com/izzet)
- [LinkedIn](https://linkedin.com/in/izzetcyildirim/)
- [Substack](https://substack.com/@izzetcyildirim)
- [ORCID](https://orcid.org/0000-0003-3513-0764)
- [X](https://x.com/izzetcyildirim)
- [Bluesky](https://bsky.app/profile/iyildirim.com)
- [Gnosis Research Center](https://grc.iit.edu/members/izzet-yildirim/)
- [DBLP](https://dblp.org/pid/360/2847.html)
- [ACM Digital Library](https://dl.acm.org/profile/99661051646)
- [IEEE Xplore](https://ieeexplore.ieee.org/author/753604940662672)
- [Semantic Scholar](https://www.semanticscholar.org/author/Izzet-Yildirim/2266402925)
- [OpenAlex](https://openalex.org/authors/A5088056256)
- [Scopus](https://www.scopus.com/authid/detail.uri?authorId=58729161800)
- [ResearchGate](https://researchgate.net/profile/Izzet-Yildirim/)
- [Wikidata](https://www.wikidata.org/wiki/Q141440172)
- [DEV](https://dev.to/izzet/)
- [SciProfiles](https://sciprofiles.com/profile/izzetcyildirim)