Hej!
Chinmay Rozekar

Can't wait to meet you.

I am Chinmay Rozekar

I build automation and agentic infrastructure that accelerates validation, debugging, and quality engineering for large, complex platforms.

2017 – 2020
RIT

Rochester Institute of Technology

MS Electrical Engineering — Graduate Researcher

  • Developed a process flow for thin-film IC fabrication, covering multiple process steps end to end.
  • Designed and simulated NMOS devices to determine sub-threshold voltages and model leakage current.
  • Fabricated a thermally actuated four-legged MEMS silicon micro-robot (SolidWorks + COMSOL Multiphysics FEA).
2019 – 2024
AMD

AMD

ESD Co-op → Product Development Engineer, System-Level Test

  • Raised silicon yield from 42% to 85% by debugging BIOS/voltage/frequency/thermal parameters across CPU and memory subsystems.
  • Built a Python + Selenium WorkOrder Automator, cutting manual work-order filing by 90%.
  • Led cross-functional debug across diagnostics, BIOS, and validation teams; built RMA/JTAG debug workflows and in-house server-farm test infrastructure.
  • Earlier, as an ESD co-op: automated Human Body Model (HBM) test programs (−90% engineering effort) and rebuilt a 20-year-old robotic GUI in Python (−95% manual interaction time).
2024 – 2025
SIEMENS

Siemens EDA

Software QA Engineer, Calibre PERC

  • Cut regression setup/debug time 20% by automating testcase checkout, config validation, and log parsing in Python/Bash/Tcl.
  • Cut nightly regression failures 15% with pre-run validation checks for environments, configs, and execution guards.
  • Validated SVRF/TVF rules across single-threaded, multi-threaded, and distributed (MTFlex) modes for Calibre PERC releases.
2025 – present

AI & Agentic Systems

triagent · punjab-pulse · applied research

  • triagent — a local, sovereign agentic RAG log-triage system (Drain3 + FAISS + Ollama) for 80GB+ semiconductor/network logs.
  • punjab-pulse — a live news aggregator (Next.js/FastAPI/Postgres) deployed on Vercel + Render + Neon.
  • Published research on high-throughput log parsing architectures and AI-integrated open-source EDA toolchains.
now

Learning OpenUSD

Why this site exists

  • This entire site — every block, every building — is authored in OpenUSD with Python (pxr), then exported through a hand-built USD→glTF pipeline for real-time web delivery.
  • Built in pursuit of NVIDIA's Omniverse platform: agentic test infrastructure applied to a complex, composition-heavy SDK.
  • Still under construction, on purpose — this chapter updates as the work continues.
2026 – Present
NVIDIA

NVIDIA

This chapter hasn't been written yet — check back once it has.

Let's talk

chinmay.rozekar.careers@gmail.com