Join us for an engaging presentation on cutting-edge research at the intersection of artificial intelligence and high-performance computing infrastructure. PhD candidate Kyrian Chinemeze Adimora will present AIMS (Adaptive Intelligent Multi-Objective Scheduler), a novel machine learning framework that optimizes supercomputer resource allocation.
What You'll Learn:
Modern supercomputers and AI datacenters face critical challenges in managing thousands of concurrent jobs across heterogeneous hardware resources (CPUs, GPUs, FPGAs). Traditional scheduling approaches optimize for single objectives (typically throughput), leading to significant energy waste, underutilized resources, and reduced system reliability. AIMS addresses these limitations through an innovative multi-objective optimization framework that simultaneously balances performance, energy efficiency, and system resilience.
Who Should Attend:
This presentation is designed for a broad audience with varying technical backgrounds:
– AI/ML enthusiasts interested in reinforcement learning, graph neural networks, and applied machine learning
– Computer science students exploring systems research, distributed computing, and infrastructure optimization
– Engineering students curious about sustainable computing and energy-efficient system design
– Aspiring researchers considering graduate studies in AI, HPC, or related fields
– Industry-minded students interested in datacenter operations, cloud computing, and enterprise-scale resource management
No prior expertise in HPC is required, the presentation uses intuitive analogies and visual explanations to make complex concepts accessible while providing technical depth for advanced students.
Speaker Bio:
Kyrian Chinemeze Adimora is a Third-year PhD candidate in Electrical Engineering and Computer Science at the University of Kansas, advised by Dr. Hongyang Sun and currently serving as the region 5 IEEE Kansas City Section Educational Activities Chair. His research focuses on applying machine learning techniques to optimize large-scale computing infrastructure. He has published at premier venues including SC ( The International Conference for High Performance Computing, Networking, Storage, and Analysis), IEEE TPDS (IEEE Transactions on Parallel and Distributed Systems) and collaborates with Argonne National Laboratory. His work bridges theoretical machine learning advances with practical systems deployment, aiming to make computing infrastructure smarter, greener, and more efficient.
Co-sponsored by: University of Kansas AI Club
Agenda:
AGENDA
5:00 PM – 5:05 PM
Welcome & Introductions
– KU AI Club president welcomes attendees
– Brief overview of AI Student Branch activities
– Presenter introduction
5:05 PM – 5:10 PM
Opening: The HPC Scheduling Crisis
– Scale of modern supercomputing challenges
– Motivation: Why this research matters
– Pizza analogy: Making the problem relatable
5:10 PM – 5:25 PM
The AIMS Solution Framework
– Three fundamental problems AIMS addresses
– Architecture overview: GNN + RL + Uncertainty Quantification
– Graph neural networks: Understanding system relationships
– Reinforcement learning: Learning from experience
– Multi-objective optimization: Balancing competing goals
5:25 PM – 5:40 PM
Results & Real-World Impact
5:40 PM – 5:45 PM
Broader Impact & Future Directions
– Applicability to cloud computing and AI datacenters
– Publications and ongoing research
– Future work: Exascale scaling, production deployment
– Career opportunities in HPC and infrastructure research
5:45 PM – 6:00 PM
Interactive Q&A Session
– Audience questions (technical and career-focused welcome)
– Discussion of implementation details
Post-Event:
– Slides and recording distributed via email
– Survey link for feedback
– Follow-up resources and reading materials
Room: 3151, Bldg: School of Engineering, Learned Hall, University of Kansas, School of Engineering, Learned Hall, Lawrence, Kansas, United States, 66045, Virtual: https://events.vtools.ieee.org/m/511042