CV

Curriculum vitae of Jun Wang.

Contact Information

Name Jun Wang
Professional Title Agent Development / Evaluation Intern at ByteDance
Email wangjun170417@163.com

Professional Summary

I build practical AI systems for coding agents, tool-use evaluation, retrieval-augmented generation, and multimodal benchmarks.

Experience

  • 2025 - 2026

    Beijing, China

    Agent Development / Evaluation Intern
    ByteDance
    • Integrated BFCLv4 into internal evaluation workflows for automated function-calling assessment.
    • Built distributed Docker sandbox execution to improve stability and isolation.
    • Designed multi-turn agent interaction and session state management for long-context tasks.
    • Reproduced VLM benchmark pipelines and implemented LLM/VLM metrics such as factuality, BON, and WON.
  • 2024 - 2024

    Beijing, China

    AI Algorithm Intern
    LargeV Instrument Corp., Ltd.
    • Applied YOLO models on oral scan datasets and improved model quality through tuning and data processing.
    • Integrated model-assisted annotation workflow to reduce manual labeling effort.

Education

  • 2026 - 2027

    San Diego, CA, USA

    Master's Program / CSE CS75 Project
    University of California, San Diego
    CSE CS75 Project
    • Expected to begin in September 2026.
  • 2022 - 2026

    Beijing, China

    Bachelor of Agriculture in Forestry; Bachelor of Engineering in Computer Science and Technology (Minor)
    Beijing Forestry University
    Forestry; Computer Science and Technology

Publications

Projects

  • Code Agent Bot

    Autonomous coding agent with planner-analyzer-critic workflow.

    • ReAct-based planning and dynamic replanning from runtime feedback.
    • MCP-aligned tool integration and chained execution with verification.
    • Critic-based self-correction for better long-task robustness.
  • Medical RAG QA System

    Retrieval-augmented QA with multi-model backend and quality evaluation.

    • Two-stage retrieval with FAISS and reranker.
    • Model-agnostic inference layer with streaming UI.
    • Automated quality evaluation with retrieval and answer metrics.
  • LoRA Forestry QA

    Domain-adapted Qwen QA model with lightweight fine-tuning.

    • Built domain QA dataset and LoRA fine-tuning workflow.
    • Added interactive streaming UI for demo and testing.
  • Multimodal Waterbird Recognition

    Image-audio recognition workflow with backend service deployment.

    • Trained multimodal models and deployed inference API.

Awards

  • 2024
    National First Prize, Multimedia Application Competition
    National-level Competition
  • 2024
    ACM Programming Competition Silver Award
    Beijing Forestry University
  • 2025
    TOEFL 103
    ETS

Skills

AI and ML (Advanced): Agent Systems, LLM Evaluation, VLM Benchmark, RAG, NLP, CV
Engineering (Advanced): Python, PyTorch, Docker, LangChain, Streamlit, Git

Languages

Chinese : Native
English : Professional working proficiency