Hi 👋, I'm Zifeng Wang

GPA: 3.89

Courses: Computer Systems, Software Construction, Distributed Systems, Speech Recognition

GPA: 3.94

Courses: C++ Programming, Advanced Mathematics, Linear Algebra, Signals and Systems, Power Electronics

• •Developed AI agent using LangChain interacting with compilation environments identifying and correcting errors in real-time.
• •Designed DFG (Data Flow Graph) Analyzer, Logic Verifier, and Test Failure Analyzer achieving 12.5% accuracy improvement.
• •Implemented multi-agent system with Syntax, Semantics, Verification, and Optimization Agent improving error detection reliability.
• •Improved debugging accuracy by 37.5% compared to baseline LLM through specialized analysis tools integrated systematically.
• •Deployed agents on AWS EC2 with Docker containers implementing CI/CD pipeline automating testing and deployment.

• •Designed tool orchestration module dynamically selecting video analysis tools based on user queries achieving intelligent routing.
• •Developed multi-modal fusion mechanism integrating text, image, and video enhancing content understanding accuracy.
• •Optimized Q&A module reducing response time improving accuracy enabling multi-turn conversational interactions reliably.

• •Designed optimization methods for maglev train system promoting performance regarding energy, safety, and passenger comfort.
• •Developed point decomposition and dynamic model construction enabling operators to solve optimal operation schemes.
• •Applied Deep Q-Network algorithm optimizing maglev system for multiple objectives reducing energy consumption by 19%.

• •Explored multimodal fusion mechanism designing intelligent perception algorithm for trains achieving safe trustworthiness.
• •Designed foreign object detection algorithm developing fusion model enhancing performance through reinforcement learning.
• •Utilized CNNs and object detection models adjusting strategies and hyperparameters for model optimization systematically.
• •Integrated RGB cameras, infrared cameras, and lidar fusing multimodal images based on attention mechanism network reliably.

• •Designed threshold detector for neural signal processing adjusting supply voltage reducing consumption achieving optimization.
• •Implemented logical effort model performing energy-delay analysis evaluating delay of critical path systematically.
• •Designed 4-bit comparator achieving 78% energy reduction and 1.5× delay increase delivering optimized circuit performance.

• •Led 4-member team exploring subway stray current impact on urban power grid transformers achieving comprehensive analysis.
• •Conducted field trips to Nanchang Metro collecting data and performing quantitative analysis on stray current paths.
• •Utilized EMD decomposition and LSTM with attention mechanism predicting current values controlling model error to 0.2%.
• •Developed transformer neutral point prediction system V1.0 and submitted software copyright and invention patent applications.