Data-Intensive Distributed Systems LaboratoryCS Seminar
Date: February 11th, 2026
Time: 12:45pm - 1:45pm
Room: SB111
Dr. Anthony Kougkas
Associate Research Professor
Deputy Director of the Gnosis Research Center
Computer Science Department
Illinois Institute of Technology
Talk Title
From Storage Optimization to Context Orchestration: Why Your AI Infrastructure is Already Obsolete
&
An Introduction to Context Engineering for Scientific Computing
Talk Abstract
The infrastructure that powers modern scientific computing was designed for humans running batch jobs, not for AI agents orchestrating complex workflows. This fundamental mismatch explains why billion-dollar AI investments often deliver zero value: context quality, not model size, determines whether autonomous systems succeed or fail.
This talk introduces context engineering as an emerging discipline for organizing the messy information that AI systems need to understand and act effectively. Through a live demonstration contrasting a simple chatbot with a sophisticated multi-agent system using orchestrator patterns, meta-prompting, and human-in-the-loop design, we will expose the hidden infrastructure that separates toy demos from production-ready autonomous science.
Drawing from the IOWarp project, a $5M NSF-funded platform bridging AI agents and high-performance computing, we will explore the four core strategies for context management: Write, Select, Compress, and Isolate. We will examine critical failure modes like context poisoning, where bad data cascades through agent workflows to corrupt scientific results, and debate whether the future belongs to general-purpose agents or domain-specific scientific systems.
This talk maps open problems in context quality metrics, provenance tracking, and evaluation frameworks, an interdisciplinary space where systems programming, AI/ML, and domain expertise converge. Attendees will leave understanding why context engineering may be the most important field they have never heard of, and how it positions AI as a research amplifier rather than a replacement for scientific intuition. The technical talk focuses on the systems engineering behind the IOWARP project, including a live demo of the agent infrastructure (https://github.com/iowarp).
Speaker Bio
Dr. Anthony Kougkas is an Associate Research Professor at the Illinois Institute of Technology and Deputy Director of the Gnosis Research Center. He also serves as Guest Research Faculty at Argonne National Laboratory. His research journey, from LABIOS (HPDC'19 Best Paper) through the Hermes I/O buffering system to the current IOWarp platform, traces the evolution from storage optimization to context engineering for AI-driven scientific computing. As Chair of IIT's University Council for AI Research Strategy, he guides research teams through practical AI deployment while leading a team of 10+ PhD students and 50+ community contributors on NSF and DOE-funded projects. His work focuses on building the infrastructure layer that enables AI agents to orchestrate scientific discovery, bridging multi-tiered storage systems with the dynamic context demands of autonomous research workflows.