Ph.D. Candidate · Northwestern University

Yunqi He

Physical-design methodology & ML-driven automation for SoC PPA.

I build quality-preserving legalization, GPU-accelerated placement, and GNN- and agent-driven engines that drive the full floorplan-to-route flow — turning research primitives into reliable, production-grade design automation.

Research

Improving SoC PPA across placement, legalization, timing, and routing — and composing reliable engines into an agentic design flow.

Physical Design & Legalization

A TwinTree quality-preserving legalizer and GPU-accelerated placement that turn continuous placements into legal, high-quality layouts at industrial scale.

Learn more →

ML for Physical Design

Graph neural networks for timing-driven placement initialization, routing-congestion prediction, and data-driven floorplanning.

Learn more →

Agentic Design Automation

An LLM-driven agent that plans and executes the full floorplan-to-route flow under unified PPA objectives, staying on physically valid states.

Learn more →
10+Peer-reviewed papers
ICCAD ’26Quality-preserving legalization
Cadence R&D internships
5Top-tier venues
ICCAD · DAC · DATE · AAAI · TODAES

Recent news

  • Jul 2026. Our paper **Quality-Preserving Legalization for Analytical Floorplanning** was accepted to **ICCAD 2026** (to appear).
  • Jun 2026. Returned to **Cadence Design Systems** as an R&D intern, building GNN-based routing congestion prediction for the Palladium (Z4) emulator.
  • Apr 2026. Our journal paper **Graph Neural Network based Initialization for Timing Driven Placement** appeared in **ACM TODAES**.
  • Jan 2026. Awarded the **Northwestern University Terminal Year Fellowship** (2026).
  • Nov 2025. Our paper **Physical-Aware eFPGA Redaction for Secure and Efficient Hardware IP Protection** was accepted to **DATE 2026**.
  • Sep 2025. Submitted a paper on eFPGA-based hardware IP protection techniques to DATE 2026.

All news & updates →

Let’s build better silicon.

I’m looking for full-time roles in physical design, design automation, and ML-for-EDA.