I2S and AI Research
I2S is at the forefront of generative artificial intelligence research, bringing together faculty and graduate researchers from multiple disciplines to advance both the foundations and real-world applications of AI. This page highlights the breadth of that work.
Graduate Research Projects
Graduate researchers across I2S and the University of Kansas are using artificial intelligence and machine learning to advance computing, communications, science, education, security and society.
The following projects represent just a few examples of AI-related research being conducted by I2S graduate students. Working alongside faculty mentors, these researchers are advancing new knowledge in generative AI, large language models, machine learning, trustworthy AI, autonomous systems and the societal impacts of emerging technologies.
Learn about I2S policies and guidelines for use of AI in research.
Machine Learning-Based Multi-Objective Optimization for HPC Workload Scheduling
Researcher: Kyrian Adimora
Faculty Lead: Hongyang Sun
This research develops a framework that uses graph neural networks and reinforcement learning to help supercomputers schedule jobs more efficiently, balancing performance, energy use and reliability simultaneously rather than optimizing for only one objective.
Scheduling Input-Sensitive Workloads with Runtime Prediction
Researcher: Sai Rithvik Gundla
Faculty Lead: Hongyang Sun
This project investigates whether improving runtime prediction accuracy leads to better high-performance computing scheduling performance using MRI-based brain segmentation jobs as a case study. Results suggest that calibration of runtime estimates can be more important than raw prediction accuracy alone.
Memory Does the Math: Hardware-Algorithm Co-Design for Accelerating AI Workloads Using Analog In-Memory Computing
Researcher: Ahsan Mojahidul
Faculty Lead: Tamzidul Hoque
This research explores analog in-memory computing, a technique that allows memory chips to perform calculations directly where data is stored. The work seeks to reduce the energy demands of AI systems while developing frameworks that account for real-world hardware limitations.
AI for Networks and Networks for AI: From Learning-Enabled Edge Systems to Real-World Testbeds
Researcher: Sravan Chintareddy
Faculty Lead: Morteza Hashemi
This work examines how AI can improve wireless and edge networks while also exploring how networks can better support distributed AI applications. Research areas include reinforcement learning, federated learning, edge inference, spectrum management and NextG experimental testbeds.
The Effects of Quantization on the J-Space
Researcher: Daniel Neugent
Faculty Lead: Jennifer Lohoefener
This project investigates how model quantization affects J-Space representations and the fitting of the Jacobian Lens.
How Do We Know When AI Is Doing a Good Job? AI Evaluation Across Indigenous Language and Classroom Contexts
Researcher: Chamisa Edmo
Faculty Lead: Jennifer Lohoefener
Drawing on Indigenous language and classroom case studies, this research examines the strengths and limitations of traditional AI evaluation methods and explores more context-sensitive approaches to assessing AI performance.
Application of AI for QA/QC in Geologic Datasets
Researcher: Bailey Srimoungchanh
Faculty Lead: Drew Davidson
This project applies AI to quality assurance and quality control processes for geologic datasets collected through StraboSpot. The system analyzes project structures and identifies opportunities to improve data quality and collaboration.
Can One Photo Be Enough? Improving Face Recognition for Biometric Watchlists
Researcher: Mashfiq Rizvee
Faculty Lead: Sumaiya Shomaji
This research introduces a probabilistic framework for biometric watchlist screening designed to support few-shot, cross-pose and dynamically updateable authentication systems.
AI-Enabled Privacy-Preserving Electrocardiogram Analysis
Researcher: Fairuz Shadmani Shishir
Faculty Lead: Sumaiya Shomaji
This project develops privacy-preserving deep learning methods for electrocardiogram analysis that maintain clinical utility while reducing the risk of exposing sensitive patient information.
AI-Enabled Atomic-Resolution Modeling of Whole Cells at Long Timescales
Researcher: Kamila Yunas
Faculty Lead: Michael Branicky
This work combines AI-predicted protein conformations with physics-based modeling to simulate whole cells at atomic resolution across biologically meaningful timescales.
Close Personal Relationships with AI
Researcher: Oluwaseun Sanwoolu
Faculty Lead: John Symons
This project explores philosophical questions surrounding human-AI relationships and examines aspects of close personal relationships that AI systems cannot replicate.
JIT Compilation Policy for ML Compilers
Researcher: Shriraj Vaidya
Faculty Lead: Prasad Kulkarni
This research focuses on machine-learning compiler policies that improve the balance between compilation cost and runtime performance.
A Tool-Agnostic Framework for Recovering Array Bounds in Stripped Binaries
Researcher: Ashish Adhikari
Faculty Lead: Prasad Kulkarni
This project combines static analysis techniques and large language models to improve recovery of stack-allocated array information from stripped software binaries.
Socially Engaging roBOtics (SEBO) Lab
Researcher: Sarah Sebo
Faculty Lead: Sarah Sebo
The SEBO Lab explores how social robots can support people in schools, homes and workplaces through tutoring, assistance and decision-support applications.
DeepPicar: A Low-Cost Autonomous Car Platform for ML Education
Researcher: Tyler Oswald
Faculty Lead: Heechul Yun
DeepPicar provides an affordable autonomous-driving platform that helps students learn machine-learning concepts through hands-on experimentation.
TinyLidarNet: 2D LiDAR-Based End-to-End Deep-Learning Model for F1TENTH Autonomous Racing
Researchers: Ethan Gao and Farrell Joswara
Faculty Lead: Heechul Yun
TinyLidarNet is an end-to-end deep neural network that uses LiDAR data to autonomously steer race vehicles in F1TENTH competitions.
When Alignment Breaks: On Inference Integrity in Text-to-Image Models
Researcher: Becky Lin
Faculty Lead: Bo Luo
Seeing Is No Longer Believing: A Security and Forensic Evaluation of Generative VideoPlatforms
Researcher: Becky Lin
Faculty Lead: Bo Luo
This research evaluates leading generative-video platforms, examining identity preservation, synthetic speech realism, watermarking, forensic detection and human perceptions of AI-generated media.
PhantomSeal: Proactive Deepfake Defense with Identity/Context Protection and Forensic Tracing
Researcher: Liangqin Ren
Faculty Lead: Bo Luo
This work investigates the risks associated with AI-generated imagery and evaluates methods for identifying synthetic content using large-scale datasets and detection techniques.
An LLM-Enabled End-to-End Attack Pipeline Against Speech-Based Machine Translation
Researcher: Junyi Zhao
Faculty Lead: Bo Luo
This project demonstrates how subtle modifications to audio inputs can manipulate speech-recognition and translation systems, creating security risks for cross-language communication technologies.
Benchmarking the Benchmarks: Evaluating Automated Safety Benchmarks for Small Language Models
Researcher: Nyam Shaik
Faculty Lead: Bo Luo
This research evaluates the effectiveness of automated safety benchmarks for small language models and identifies limitations in current approaches to assessing model safety.
The Adversarial AI-Art: Understanding, Generation, Detection and Benchmarking
Researcher: Yuying Li
Faculty Lead: Bo Luo
AI Data Centers in Africa
Researcher: Oluwakorede Ajibona
Faculty Lead: John Symons
This project explores the environmental implications of Africa's growing data-center infrastructure through the creation of an interactive database documenting facilities and their resource impacts.
AI Research Center Spotlight: Center for Cyber-Social Dynamics
The Center for Cyber-Social Dynamics (CCSD) advances research at the intersection of artificial intelligence, society and social change. The center pursues two complementary lines of inquiry: understanding the societal impacts of AI and developing computational methods that can detect and forecast shifts in social norms.
A flagship area of research focuses on creating AI- and machine-learning-based models capable of identifying early signals of normative change and anticipating how collective values, expectations and behaviors evolve over time. This forecasting work is grounded in, and informed by, the center's sustained scholarship on the social and ethical dimensions of AI.
CCSD researchers have published extensively in areas including value alignment theory, the use of AI in courts and judicial decision-making, AI governance and policy, the integration of AI in education, and the epistemic role AI increasingly plays in scientific inquiry.