About me

A skilled ML engineer with 4 years of experience in Machine Learning, Computer Vision, NLP, and Software Engineering. Proficient in Python, C++, PyTorch, and LLMs, with a strong track record in developing AI solutions for Object Detection, Segmentation, and RAG. Also equipped with full stack development capabilities, excelling in both frontend and backend technologies, and capable of independently managing the entire lifecycle from system architecture design to product deployment.

What I'm doing

  • AI Development icon

    AI Development

    Professional development of AI solutions including computer vision and LLM applications.

  • Software Development icon

    Software Development

    High-quality development of software applications using Python, C++, and other technologies.

  • Web Development icon

    Web Development

    Full-stack web development using modern technologies like Vue, Node.js and cloud services.

Certifications

Technical Skills

  • Programming Languages

    • Python
    • C++
    • JavaScript
    • SQL
    • Vue

  • Frameworks & Libraries

    • PyTorch
    • TensorFlow
    • OpenCV
    • LangChain
    • YOLOv11

  • Tools & Technologies

    • Docker
    • Git
    • Linux
    • Cloud Services
    • RESTful APIs

Resume

Education

  1. Chengdu University of Technology

    Sep. 2020 — Jul. 2023

    Master's in Software Engineering

  2. Changsha University of Science and Technology

    Sep. 2015 — Jul. 2019

    Bachelor's in Vehicle Engineering

Experience

  1. Grsofter - CN (Remote)

    Jan. 2025 - Present

    1. Intelligent Dialogue System for Embodied Agents:
    • Extended the open-source xiaozhi-esp32-server backend (Python + WebSocket) to support embodied interaction scenarios
    • Implemented context-aware multi-turn dialogue using LLMs with intent recognition and function-call capability

  2. MegaDairy Co - NZ

    Mar. 2024 - Jul. 2024

    1. Knowledge Base Question-Answering System:
    • Designed and implemented an intelligent Q&A system using LangChain-Chachat, improving information retrieval efficiency
    • Single-handedly managed entire project lifecycle, from system analysis and architecture design to core functionality implementation
    • Reduced internal document retrieval time from 3 minutes to under 3 seconds

    2. Inventory Management & Procurement Alert System:
    • Developed a Python-based solution to integrate multiple Excel databases (inventory records, sales orders, suppliers)
    • Implemented automated email notifications to the purchasing team with detailed shortage reports and supplier suggestions
    • Reduced manual checking time by 80% and helped prevent stockouts through proactive alerts, supporting seamless production planning

    3. PDF Label Template Generation Software:
    • Developed Python-based automation software for generating PDF inspection templates for product labels
    • Implemented image processing, text recognition, and automated naming/saving functions
    • Reduced template generation time from 5 minutes to 30 seconds per item, processing over 2,000 templates monthly

  3. TENGDEN - CN

    Jul. 2023 - Feb. 2024

    1. Airport Runway Segmentation Algorithm:
    • Developed and implemented an EGE-Unet model for high-precision airport runway segmentation, achieving an initial mIoU of 0.971
    • Enhanced model performance through advanced data augmentation techniques, resulting in an improved mIoU of 0.978
    • Optimized the algorithm for production by porting it to C++, reducing inference time by 20% and enabling real-time processing

    2. Web-based Object Detection Data Annotation Tool:
    • Created a web-based data annotation tool for object detection using JavaScript, HTML, and CSS
    • Implemented features for bounding box drawing, label management, and dataset organization
    • Improved annotation efficiency by 40% through intuitive UI design and keyboard shortcuts

  4. Sobey Digital - CN

    May. 2022 - Dec. 2022

    1. Intelligent Editing-Scene Content Recognition:
    • Achieved mAP@0.5 of 93.50% for scene content recognition using YOLOv5s

    2. Intelligent Editing-Specific Objects Recognition:
    • Fine-tuned YOLOv5m on Object365 dataset for specific object recognition, mAP@0.5 of 55.50%

  5. Wangsi Tech - CN

    Jul. 2021 - Apr. 2022

    1. Optical Character Recognition (OCR) for IP Addresses in Images:
    • Implemented accurate text recognition using PaddleOCR, and wrote scripts to filter key information
    • Developed image preprocessing techniques to improve recognition accuracy, and finally packaged it into a Docker image for submission

My skills

  • Computer Vision
    90%
  • Deep Learning
    85%
  • Software Development
    80%
  • Large Language Model
    75%

Portfolio

Blog

Contact

Contact Form