Murali Krishna Emani
Scholar

Murali Krishna Emani

Google Scholar ID: A1AoXkYAAAAJ
Computer Scientist at Argonne National Laboratory
AI AcceleratorsHigh Performance ComputingBenchmarkingRuntime SystemsParallel Programming models
Citations & Impact
All-time
Citations
1,499
 
H-index
20
 
i10-index
31
 
Publications
20
 
Co-authors
10
list available
Contact
No contact links provided.
Publications
20 items
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
  • September 2025: Paper 'AERIS: Argonne Earth Systems Model for Reliable and Skillful Predictions' selected as a finalist for the 2025 ACM Gordon Bell Prize in Climate Modeling
  • September 2025: Paper 'MoE-Inference-Bench' accepted at PMBS workshop at SC25
  • September 2025: Paper on unstructured sparse fine-tuning on wafer-scale engine accepted at ExHetAI workshop at SC25
  • July 2025: Paper 'MoPEQ' accepted at Workshop on Extreme Quantization for Computer Vision (ICCV 2025)
  • June 2025: Paper 'Langvision-Lora-NAS' accepted at ICIP 2025
  • February 2025: Papers 'AILuminate' and 'BaKlaVa' posted on arXiv
  • June 2025: Co-organized two BoF sessions and a tutorial on programming novel AI accelerators at ISC25
  • Served on program committees for top-tier conferences (e.g., SIGKDD, AAAI, SC, IPDPS) and as reviewer for journals
  • Mentored multiple postdocs and PhD students, many now at RedHat, PNNL, Nvidia, Microsoft, etc.
Research Experience
  • Currently a Computer Scientist at ALCF, Argonne National Laboratory
  • Previously a Postdoctoral Research Staff Member at Lawrence Livermore National Laboratory
  • Co-leads the AI Testbed at ALCF, exploring performance and efficiency of AI accelerators for scientific ML applications
  • Served as co-chair of the MLPerf HPC group at MLCommons for benchmarking large-scale ML on HPC systems
  • Current projects include: developing performance models for scaling ML/DL frameworks on emerging supercomputers; co-design of hardware and ML algorithms; benchmarking ML/DL methods on HPC systems