Qihui Pan

• Participate in data engineering projects for district-level courts across Guangzhou, including the cleaning, classification, standardization and structuring of judicial and case-related data to improve data quality and usability.

• Contributed to development of AI-assisted judgment document system, transforming court workflows and legal-document requirements into practical AI functions for processing case materials and supporting judgment-document preparation.

• Assist with legal-text processing, key information extraction and structured data preparation for AI applications, while participating in system testing and iterative optimization to improve the accuracy, stability and practical usability of AI-assisted outputs.

• Work with technical and project teams on real-world judicial digitalization requirements, gaining practical experience in legal AI, judicial data engineering and enterprise AI system development.

• Developed a batch enterprise credit-information query tool supporting thousands of records, multi-source APIs, asynchronous queries, retries, logging and result export.

• Evaluated and integrated Qwen-series large language models across 30+ business scenarios, comparing accuracy, response speed, hallucination control and API cost.

• Developed automated short-video merging with FFmpeg and built an academic-paper retrieval workflow covering paper filtering, PDF acquisition and key-conclusion extraction.

• Strengthened practical skills in Python asynchronous programming, LLM evaluation, FFmpeg processing and automated workflow design.

• Gained hands-on experience in various software programs, increasing proficiency and expanding technical skill set.

• Classified, filtered and structured multi-client material data and built a local machine-learning environment for model pre-annotation, training and deployment.

• Participated in automated sorting equipment testing, proposed engineering improvements, and developed a Python batch file-renaming tool to reduce manual data-processing time.

• Optimised a Nike insole 3D scan in Unity through mesh reconstruction and denoising, and implemented geometry-shader processing and a Dijkstra-based geodesic-distance experiment.

• Participated in testing and debugging software applications to ensure functionality and user experience.