I am a member of technical staff at xAI working on inference for Grok models.

I graduated with PhD in Computer Engineering from UIUC advised by Prof. Ravishankar Iyer. My thesis focused on improving the reliability and efficiency of large-scale ML systems. During my PhD, I was fortunate to collaborate with amazing mentors and productionize my research at IBM, Google, and Meta.

Selected Publications

• Characterizing GPU resilience in modern AI systems.

  Shengkun Cui^, Archit Patke^, Hung Ngyuen, Aditya Ranjan, Ziheng Chen, Phuong Cao, Brett Bode, Gregory Bauer, Catello Di Martino, Saurabh Jha, Chandra Narayanaswami, Daby Sow, Zbigniew Kalbarczyk, Ravishankar Iyer.

  SC 2025.

  • ^ Equal Contribution

  •  DOI
  •  Paper

• Learned Page Migration in Disaggregated HPC Systems.

  Archit Patke, Chrisitan Pinto, Saurabh Jha, Haoran Qiu, Zbigniew Kalbarczyk, and Ravishankar Iyer.

  ICS 2025.

  •  Paper

• QLM: Queue Management for SLO-oriented Large Language Model Serving.

  Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu, Christian Pinto, Chandra Narayanaswami, Zbigniew Kalbarczyk, and Ravishankar Iyer.

  SoCC 2024.

  • Adopted by vLLM (release notes) and ByteDance AIBrix (blog)

  • Media Coverage: IBM, ByteDance, Hugging Face

  •  DOI
  •  Code
  •  Paper

• Power-aware Deep Learning Model Serving with μ-Serve.

  Haoran Qiu, Weichao Mao, Archit Patke, Shengkun Cui, Saurabh Jha, Chen Wang, Hubertus Franke, Zbigniew Kalbarczyk, Tamer Başar, Ravishankar Iyer,.

  ATC 2024.

  •  DOI
  •  Paper

• Delay Sensitivity-driven Congestion Mitigation for HPC Systems.

  Archit Patke, Saurabh Jha, Haoran Qiu, Jim Brandt, Ann Gentile, Joe Greenseid, Zbigniew Kalbarczyk, and Ravishankar Iyer.

  ICS 2021.

  •  DOI
  •  Talk
  •  Paper

• Measuring Congestion in High-Performance Datacenter Interconnects.

  Saurabh Jha, Archit Patke, Jim Brandt, Ann Gentile, Benjmain Lim, Mike Showerman, Greg Bauer, Larry Kaplan, Zbigniew Kalbarczyk, William Kramer and Ravishankar Iyer.

  NSDI 2020.

  •  DOI
  •  Data
  •  Code
  •  Talk
  •  Paper