Hey! I am Grant Wilkins, an Electrical Engineering PhD student at Stanford University from Kingsport, TN. As AI drives unprecedented growth in electricity demand, I study how datacenters can share the grid with everyone else who depends on it by taking a computing and energy blended approach to problem solving. Some of the work we have done is in building load models, electrical prototypes, formulations and power-system tools so that their expansion stays reliable and clear rather than overbuilt and offloaded.

I am always happy to chat, so reach out to gfw at stanford dot edu for any inquiries or questions.

A few things I am or have been a part of:

• Developing grid and datacenter interaction models and tools at Stanford University
• Mitigating transients and load swings in AI training clusters at Stanford University
• Designing next-generation datacenter power hierarchies to reduce overbuilding Microsoft Azure Research–Systems
• Energy-optimizing serving and routing of LLM inference systems supervised by Richard Mortier and Srinivasan Keshav at the University of Cambridge.
• Modeling high-performance computing energy consumption supervised by Jon Calhoun at Clemson University.
• Developing lossy compression techniques for federated learning supervised by Sheng Di at Argonne National Lab.
• Designing IoT software for smart grid applications at Tesla, Inc..

Find a fairly recent CV here

Publications

Preprints

1. G. Wilkins, F. Kazhamiaka, R. Rajagopal. “From Servers to Sites: Compositional Power Trace Generation of LLM Inference for Infrastructure Planning”. 2026. arxiv

Conference Proceedings

1. G. Wilkins, S. Di, J. C. Calhoun, R. Underwood, and F. Cappello, “To Compress or Not To Compress: Energy Trade-Offs and Benefits of Lossy Compressed I/O,” in 2025 IEEE International Parallel and Distributed Programming Symposium, Jun. 2025. find paper here arxiv
2. G. Wilkins, S. Keshav, and R. Mortier. “Towards Energy-Optimal LLM Serving: Workload-Based Energy Models for LLM Inference on Heterogeneous Systems,” 2024 ACM HotCarbon Workshop on Sustainable Computer Systems (HotCarbon’24), Jul. 2024. find paper here arxiv
3. G. Wilkins, S. Keshav, and R. Mortier. “Hybrid heterogeneous clusters can lower the energy consumption of LLM inference workloads,” in 2024 ACM International Conference on Future and Sustainable Energy Systems (e-Energy ‘24), Jun. 2024, pp. 506–513. find paper here arxiv
4. G. Wilkins, S. Di, J. C. Calhoun, K. Kim, R. Underwood, and F. Cappello, “FedSZ: Leveraging error-bounded lossy compression for federated learning communications”, in 2024 IEEE International Conference on Distributed Computing Systems (ICDCS), Jul. 2024. find paper here arxiv
5. G. Wilkins, M. J. Gossman, B. Nicolae, M. C. Smith, and J. C. Calhoun, “Analyzing the energy consumption of synchronous and asynchronous checkpointing strategies”, in 2022 IEEE/ACM Third International Symposium on Checkpointing for Supercomputing (SuperCheck), Nov. 2022, pp. 1–9. find paper here
6. G. Wilkins and J. C. Calhoun, “Modeling power consumption of lossy compressed i/o for exascale hpc systems”, in 2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), Jun. 2022, pp. 1118–1126. find paper here
7. G. Wilkins and J. C. Calhoun, “Modeling energy Consumption for the SZ compressor on hpc systems”, in IEEE/ACM 32nd International Conference for High Performance Computing, Networking, Storage, and Analysis Proceedings, Oct. 2020. find paper here

Journal Articles

1. S. Di, J. Liu, K. Zhao, X. Liang, R. Underwood, D. Tao, J. Tian, Y. Huang, J. Huang, X. Yu, J. C. Cahoun, M. Shah, B. Zhang, G. Wilkins, Z. Zhang, G. Li, K. A. Alharthi, and F. Cappello, “A survey on error-bounded lossy compression for parallel and distributed use-cases”, ACM Computing Surveys, In Submission, 2024. arxiv

Theses and Reports

1. G. Wilkins, “Online Workload Allocation and Energy Optimization in Large Language Model Inference Systems”, University of Cambridge MPhil in Advanced Computer Science, June 2024. find paper here
2. G. Wilkins, “Green HPC: Optimizing Software Stack Energy Efficiency of Large Data Systems”, Clemson University Honors College, May 2023. find paper here

Experience

Research

1. Graduate Researcher at Stanford University (Fall 2024 to present)
   Advisors: Ram Rajagopal, Phil Levis
   Lead on developing datacenter level load models to integrate into grid planning methodology.
   Assistant on designing power electronics solution to swings in AI training power draw.

2. Research Intern at Microsoft Azure Research–Systems (Summer 2025)
   Advisors: Fiodar Kazhamiaka, Alok Kumbhare, Chaojie Zhang, and Ricardo Bianchini
   Created datacenter power hierarchy simulator to explore reliable and efficient designs for rising power density deployments.

3. Graduate Student Researcher at Argonne National Laboratory (Summer 2023 and 2024)
   Advisors: Sheng Di, Robert Underwood, and Franck Cappello
   Developer of FedSZ: a lossy compressor to reduce the overhead of federated learning communications.
   Lead on a study to quantify energy benefits of using lossy compression to cut data size.

4. Graduate Student Researcher at University of Cambridge (Fall 2023 to Summer 2024)
   Advisors: Richard Mortier and Srinivasan Keshav
   Developer of EASLI: an online, energy-aware scheduler for serving LLM inference.

5. Undergraduate Researcher at Clemson University (Fall 2020 – Fall 2023)
   Advisor: Jon Calhoun
   Worked on projects concerning green supercomputing with focuses on optimizing the energy consumption of lossy compression, checkpoint-restart, and heterogeneous computing.

6. NSF-REU: HPC Data Reduction at Clemson University (Summer 2020)
   Advisor: Jon Calhoun
   Produced runtime power models of the SZ lossy compressor for optimization.

Industry

1. Software Engineering Intern. Tesla, Cloud Energy Platforms Team (Summer 2021)
   Worked on California Virtual Power Plant, a project that ended up keeping power on for thousands during a wildfire.
   

Teaching

Clemson University ENGR 1410 Introduction to MATLAB.
Teaching Assistant, Fall 2020, Spring 2021.

(Last update: Mar 24, 2026)