Drimva

NUVACORE

CPU Performance Modeling (Lead & ICs)

$200K–$300K

United States +3 · Hybrid · Full-time · Posted Sep 22, 2026

Location
United States
Canada
India
Remote
Experience
20+ years (Lead) or 8+ years (ICs) in CPU microarchitecture and/or performance engineering, and the same years of hands-on experience with cycle-accurate C++ CPU performance simulators.
Education
MS in Computer Architecture, Computer Engineering, or related field (PhD preferred).

About the job

Lead or contribute to CPU performance modeling and microarchitectural exploration, using cycle-accurate simulators to guide design decisions from early pathfinding through tape-out and post-silicon correlation.

What you'll do

  • Own the cycle-accurate CPU performance simulator and modeling methodology.
  • Explore CPU microarchitecture and debug performance miscorrelation across the model, RTL, and post-silicon results.
  • Collaborate across design and software teams, mentor engineers, and contribute to patents or publications where applicable.

What they ask for

  • MS in Computer Architecture, Computer Engineering, or a related field.
  • Hands-on experience with cycle-accurate C++ CPU performance simulators; strong C++ and Python / Perl skills, and ability to write and analyze assembly.
  • Deep expertise in at least one listed CPU domain; experience in RISC-V or ARM64 or x86 ISA.
  • Track record driving performance from pathfinding through silicon, improving CPU PPA, and developing trace-based and execution-driven simulators.

Nice to have

  • PhD.
  • Compiler optimizations and hardware/software co-design; published work or filed patents in CPU performance or microarchitecture; prior principal-level role at a CPU design organization.

Pay

USD 200,000–300,000 (per-year-salary) base salary

Tools

C++, Python, Perl, gem5, RISC-V, ARM64, x86

AI focus

The CPU is intended for AI infrastructure and agentic computing workloads; the role models and explores CPU performance for these demands.

Model-written summary · the original posting is authoritative

First seen Oct 29, 2026 · Last verified Nov 5, 2026 (UTC)