Prachatos Mitra’s homepage
Hello, I am a PhD student at Georgia Tech’s School of Computer Science, advised by Prof. Alexandros Daglis. My interests are in computer architecture, particularly on the evaluation of emerging systems at massive scale. Currently I work on performance tuning for actor-model workloads on commodity hardware, and on the design and evaluation of a large-scale actor-model-based graph processing system, where I build methods for modeling workload behaviour at scales too large to run directly.
Prior to this, I was a Senior Silicon Design Engineer at AMD India, working in the Server Performance Group. I worked on the AHV hypervisor at Nutanix India for two years. I received my Master’s degree from the Indian Institute of Science (IISc) in 2020. At IISc, I was a part of the Computer Systems Lab with Prof. Arkaprava Basu. I worked on new TLB and cache replacement policies based on dead-entry prediction [HPCA ‘21] at IISc.
Recent work
MacroPattern captures how a workload behaves as a small set of fitted distributions per handler, scales those fits analytically to the target problem and system size, and feeds them to MPSim, a parallel discrete-event simulator that reaches 16 million processing elements.
Projected message rate against problem scale for four graph workloads. Beyond the training scales, MacroPattern stays on the empirical curve where naive scaling and linear regression diverge, because it carries the message distributions forward rather than extrapolating a trend.
I am looking for research and performance modeling internships for Summer 2027.