|
I am a senior undergraduate majoring in Software Engineering at Wuhan University (National Characteristic Demonstration Software College). My passion lies in bridging the gap between Academic Algorithms and Production Engineering. I specialize in designing robust Multi-Agent Systems and optimizing Embedded AI pipelines. Currently, I am conducting research on 2D-3D Person Re-Identification (ReID) targeting IJCAI submission.
|
  -yellow?style=social&logo=authy){kind=icon} 🏆 Top Awards:
|
| Languages | AI & Data Science | Embedded & IoT | Tools & Dev |
|---|---|---|---|
   |
   |
   |
   |
Detailed breakdown of my key engineering challenges and solutions.
- Role: Core Architect & Developer (48-hour Hackathon)
- The Problem: The native platform lacked state memory, couldn't handle "Intent Uncertainty" (IIUN), and failed in multi-turn interactions.
- My Solution - A4R (Ask-for-Refine):
- Intelligent Pre-processing: Designed a trio of agents:
GuideAgent: Entry point for traffic routing.ValidateAgent: Embedded QA validation to fix the multi-turn interaction defect.DiscernAgent: Converts fuzzy user intents into actionable tasks.
- Engineering Resilience:
- Implemented a "Task Fingerprint" algorithm to deduplicate tasks and prevent redundant LLM calls.
- Built a "LLM + Rule Engine" dual-track strategy to ensure availability even if the API fails.
- Achieved full-link tracing for 8 concurrent agents using unified context IDs.
- Intelligent Pre-processing: Designed a trio of agents:
- Role: System Lead (Logic & Algorithms)
- Tech Stack: Python Threading, LLM Prompt Engineering, Audio Processing.
- Key Technical Contributions:
- Concurrency Control: Solved critical conflicts between "Real-time Debugging" and "Hardware Control" using a robust Python
threadingarchitecture. - Prompt Engineering: Iterated prompts through 3 versions to boost user intent understanding accuracy to 92%.
- Performance: Optimized the audio end-to-end pipeline, achieving a response time of < 1.5 seconds.
- Concurrency Control: Solved critical conflicts between "Real-time Debugging" and "Hardware Control" using a robust Python
- Focus: Dynamic Orchestration & Collaborative Reasoning.
- Contributions:
- Developed key modules for Agent Workflow, specifically tool calling and decision logic.
- Optimized Inter-Agent Communication Protocols to enhance robustness in multi-step complex tasks.
- Tech: RISC-V, ESP32, Sensor Fusion.
- Achievement: Independently developed underlying drivers for a dual-architecture system (RISC-V + ESP32), adapting 6 types of sensors.
- Impact: Reduced the learning curve for students by 40% during pilot testing in experimental classes.
-
Head of Campus News Center (Student Org)
- Led a team to publish 100+ articles (including 17 deep-theory pieces), achieving 1000+ avg. views/article.
- Implemented standardized management workflows, achieving zero errors throughout the year.
-
Campus Activity Platform QA (Student Union)
- Responsible for platform quality assurance and bug tracking.
- Identified and fixed 15+ critical bugs, improving system stability by 30%.
Let's connect and build something amazing! 🚀
"Code is not just about instructions; it's about solving reality's uncertainty."
Last Updated: 2026