Pin-Han (Peter) Chen

Southern California Computer-Aided Design Lab (SCCAD)

Graduate Research Assistant

- Developing end-to-end 2.5D IC design workflow for mechanical integrity using Siemens and Ansys EDA tools
- Building ML framework for foundation model pretraining in 2D IC global routing
- Targeting multi-objective RL model with differentiable physics modeling for 2.5D/3D ICs

MediaTek

Research Scientist, Strategic Technology Team

- Built multi-agent workflow for RF/Analog Design using AutoGen, outperforming benchmark by 39%
- Invented Convolutional KAN for Power Amplifier modeling with 2.2% improvement and 40% fewer parameters
- Developed physics-aware DRL model for inverse CoWoS design, saving 60% simulation time

Johns Hopkins Hospital

Machine Learning Intern

- Led a team of 5 interns building apeak-time prediction model achieving 95% accuracy
- Enhanced feature engineering by designing a sequential data pipeline with 170k patient dataset
- Invented two ensemble and weighted algorithms, leading to a 12% improvement in model performance and efficiency

MediaTek

Machine Learning Engineer, AI Team

- Achieved top 1% ranking among entry-level employees, earning two distinguished job performance awards
- Developed a DRL model for macro placement optimization, alleviating the dependency on rule-based search
- Implemented pre-trained/fine-tuned pipelines with MLflow, saving 50% of training costs

Computer Games and Intelligence (CGI) Lab

Research Assistant

- Created DRL model with PyTorch and Cython achieving 3x speedup
- Initiated first MuZero-based model research on Chinese Dark Chess
- Boosted model performance by 15% through randomness conceptualization neural network