Selected Projects & Research

🏆 NSTC Research Award Winner

• Engineered a high-performance, GPU-accelerated global router for large-scale industrial designs with over 200 million pins.
• Implemented a hybrid of Reinforcement Learning (RL) and rule-based policies to make intelligent, timing-driven routing decisions.
• Outperformed state-of-the-art routers on multiple industry benchmarks, achieving a 16x computational speedup against GAMER and improving critical timing (WNS) by over 55% against Instant GR.

🥇 Global 1st Place at IEEE/ACM CAD Contest

• Architected a scalable, true-3D analytical placer for large-scale, F2F stacked ICs, proven on complex industrial designs with 740k+ cells.
• Utilized a multi-technology objective function and HBT-cell co-optimization to directly optimize for the final PPA.
• Delivered a state-of-the-art placer that achieved 6.4% performance gains over the 2nd-place team.

• Developed a novel, end-to-end RLAIF pipeline for voice conversion, leveraging the Hugging Face TRL library to fine-tune transformer models with Direct Preference Optimization (DPO).
• Engineered a multi-objective reward function by fusing CLAPS, ASR, and MOS signals to improve alignment and vocal fidelity.

• Engineered a complete autonomous navigation system on ESP32 and PSoC microcontrollers, using ultrasonic sensors for dynamic obstacle avoidance and a fine-tuned PID controller for precise line tracking.
• Architected the underlying embedded system with FreeRTOS for real-time performance and utilized AWS IoT for remote dispatch.

• Architected a full-stack platform serving 1,000+ users with Node.js RESTful API, JWT authentication, React frontend, real-time WebSocket chat, and MongoDB database.
• Built a CI/CD pipeline with Docker for automated GCP deployment, and reduced database query latency by 40% via indexing.