Max Hawkins

CSE PhD student at Georgia Tech proposing a general measure of computational performance

I’m a Computational Science and Engineering PhD student at Georgia Tech exploring general measures of computational performance under the guidance of Richard Vuduc.

Thesis Research: Information Theory for Computing Performance Measurement

My primary work involves exploring the use of information theory to evaluate computing performance more generally than current metrics (e.g. flop/s). The growing diversity of data types and emulation strategies shows that the standard unit of work, an IEEE-754 FP64 addition or multiplication, is too narrow for modern hardware. Going further, how can we generalize computational work across not just datatype bit-widths and emulation schemes but also across sparsity, noise, and analog, neuromorphic, or quantum computing?

Here’s a 4-page preprint containing the main gist of my proposal: Back to Bits: Extending Shannon’s Communication Performance Framework to Computing.

I also wrote a poem to concisely capture the ideas of my PhD thesis through rhythmic verses: Farewell to the Flop - A Post-Dennard Poem.

Extending this work to the classic roofline model results in the information roofline. Here is a two-page extended abstract of what will later be a full paper: Information Roofline: A Model for the Era of Datatype Optimization and Emulation.

Applying my thesis to datatype design was discussed at the Conference for Next-Generation Arithmetic in 2025: Talk Slides.

I gave a presentation on applying this work to noise and sparsity at the Neuro Inspired Computational Elements Conference (NICE) in 2026: Talk Slides, Talk Video.

Other Research Interests

Societal Impacts of Computing as a Physical Process

Computation isn’t some ethereal process. It’s a real, physical process that requires energy and often water. The large-scale adoption of AI has immense downstream effects on our energy grid, water resources, and society in general. More work needs to be done to quantify and reduce these effects. Specifically, I’m working on projects related to datacenter-energy grid cooperation, datacenter design and integration into urban design, and accurately quantifying and conveying a user’s HPC energy/water/carbon footprint.

Scientific Dissemination, Curation, and Peer Review

The modern scientific process is clearly under stress. Due to publish-or-perish pressures, adoption of AI workflows, and many other factors, the number of academic papers submitted is growing beyond what our current systems can handle. What does the future of academia look like at scale? How can we maintain rigorous peer review? How do we move towards freely accessible scientific publications?

I have a prototype infrastructure software package I call Science Cove that intends to help make a better path forward. Coming soon…

Teaching

I enjoy teaching and have served as a TA for several semesters. Through Georgia Tech’s OMSCS Pre-Doctoral Fellowship, I created and taught a seminar on datacenter design, operation, and societal impact and mentored 12 OMSCS students in research. Public course materials are here: Computing at Scale.

  • Fall 2026 - OMSCS High-Performance Computing
  • Summer 2026 - OMSCS High-Performance Computing
  • Spring 2026
    • Computing at Scale: The Design, Operation, and Societal Impacts of Datacenters
    • Research Seminar for 12 OMSCS students
  • Fall 2025 - Computing at Scale: The Design, Operation, and Societal Impacts of Datacenters
  • Summer 2025 - OMSCS High-Performance Computing
  • Spring 2025 - OMSCS High-Performance Computing

Publications and Presentations

  • ModSim ‘26, “Information Roofline: A Model for the Era of Datatype Optimization and Emulation”, Max Hawkins and Rich Vuduc, Extended Abstract, Poster

  • Neuro Inspired Computational Elements Conference (NICE) ‘26, “Bridging Neuromorphic and Traditional Computing Performance: An Information-Theoretic Approach”, Max Hawkins and Richard Vuduc, Talk Slides, Talk Video

  • Reed College Colloquium ‘26, “Reframing Computers as Information Channels”, Max Hawkins

  • Computational Mathematics (COMA) Seminar, Technical University of Munich ‘25, “Back to Bits: Extending Shannon’s Framework for Communication Performance to Computation”, Max Hawkins, Invited Talk (Virtual)

  • HotCSE Seminar ‘25, “Back to Bits: Extending Shannon’s Framework for Communication Performance to Computation”, Max Hawkins, Talk Slides

  • Conference for Next-Generation Arithmetic ‘25, “Bits on Bits: Showcasing Next-Gen Arithmetic through Information Theory”, Max Hawkins and Rich Vuduc, Talk Slides

  • arXiv preprint ‘25, “Back to Bits: Extending Shannon’s communication performance framework to computing”, Max Hawkins and Richard Vuduc, Preprint

  • International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences (ISPRS) ‘25, “A Data-driven Smart District Toward Net Zero Using Generative Design and Urban Digital Twins: A Use Case of Nihonbashi, Tokyo”, Alorabi, Shen, Smith, Hawkins, et al., DOI, Paper

  • ModSim ‘25, “AI Datacenter Power Curtailment - Potential, Challenges, and Implementation”, Max Hawkins and Richard Rex, Poster, Slides

  • Energy HPC Conference ‘25, “Runtime and Energy Analysis of SpMV Hardware Execution Choice”, Max Hawkins, Christian Engman, and Ivan Rocha, Talk Video, Slides

  • JuliaCon ‘22, “Metal.jl - A GPU backend for Apple hardware”, Max Hawkins and Tim Besard, Talk Video, Code

  • URSI GASS ‘21, “High-Performance Radio Telescope Array Data Processing Framework”, M. W. Hawkins, D. J. Czech, D. H. E. MacMahon, S. Croft, and A. P. V. Siemion, Paper

  • Council on Undergraduate Research Posters on the Hill ‘21, “High-Performance Radio Telescope Array Data Processing Framework”, M. W. Hawkins

  • AAS 235 ‘20, “High Time-Resolution Radio Frequency Interference and Single Pulse Pulsar and FRB Detection using Machine Learning Semantic Segmentation”, Hawkins, Lynch, Hawkins, and Smith

  • AAS 235 ‘20, “Ultra-wideband Digital Technologies for the Green Bank Telescope”, Lynch, Hawkins, McCullough, Ray, Jensen, Hawkins, et al.

Contact and Personal

If you have a comment, question, or otherwise want to contact me, please email mhawkins60@gatech.edu. I’m also open to mentoring/collaborating with both undergraduate and graduate students in any of the above areas.

Personally, I enjoy working with my hands and touching grass when I’m not studying or coding. The implementations of that idea are seen in my interests in pottery, woodworking, gardening and poetry. I also love to travel and try to keep myself open to unexpected, novel experiences.

Last Updated: August 31st, 2026