Tian, Qijia

EDUCATION

• Msc in Computer Science, HKU, with distinction, GPA: 3.58/4.3;
• Courses: Natural Language Processing (A); Introduction to Unmanned Systems (A-); Big Data Management (A-); Introduction of Cyber Securities (A).

• BENG in Computer Engineering, HKU, with First Class Honor, GPA: 3.77/4.3, Rank: 1/52;
• Courses: Senior Design Project (A-); Introduction to Deep Learning for Computer Vision (A+); Algorithm Design (A); Digital System Design (A+); Computer Architecture (A); Control System I (A+); Probabilistic System Analysis (A+); Signal and Linear Systems (A-).

EXPERIENCE

• Designed a reproducible Medical OCR Benchmark for clinical intake forms, focusing on robust extraction from noisy handwritten and scanned fields;
• Established a two-track OCR protocol: V1 (full-text transcription) and V2 (handwriting + Y/N key fields); benchmarked 10+ vision-language models on 49 anonymized cases, with best results at CER 3.67% / WER 6.63% and Y/N accuracy up to 94.94%;
• Currently advancing a triage + decision + audit pipeline with evidence retrieval and cross-checking modules, so each recommendation is grounded in source records and model outputs;
• Implementing verifiable and traceable workflow logging (input snapshots, intermediate reasoning artifacts, and decision provenance) to support clinical review, audit, and model governance.

• Contributed to deployment of a Python-based TSN verification platform to evaluate Delay Alignment Scheduling Algorithm (DASA) on industrial ring topologies;
• Processed and visualized large-scale simulation results, validating 36% reduction in maximum delay and 81.9% reduction in delay difference versus baseline FRER methods;
• Supported literature review on TSN reliability mechanisms and assisted manuscript formatting and camera-ready preparation.

• Tested systems with Jmeter/Python and supported issue fixing and platform optimization.

• Supported motor design/testing and quality assurance to meet industry standards.

PROJECT EXPERIENCE

• Architected and implemented high-performance Entity Component System (ECS) with archetype-based entity management, supporting 10,000+ entities at 100+ FPS;
• Designed efficient ECS query and iteration mechanisms optimized for cache performance and scalability;
• Developed comprehensive sample game demonstration (3D vampire survivor style) showcasing engine capabilities;

• Developed a modular AI research assistant framework enhancing OS-Copilot with script-based architecture;
• Implemented hierarchical documentation system combining high-level tool menus and detailed guides;
• Created platform integrations for Hugging Face, GitHub, and X (Twitter) for automated data retrieval;

• Enhanced model interpretability using Grad-CAM, SHAP, LIME, and Occlusion Sensitivity Map techniques for breast tumor segmentation;
• Implemented Attention UNet for accurate segmentation under computational constraints;
• Developed surgical robot hardware prototype with motor control and real-time processing;

• Developed an AI system to streamline petition management with data analytics and automation;
• Classified over 700 types of petition work orders with high accuracy;
• Integrated data from petition records, historical cases, and policy data to enhance workflow;
• Played a key role in algorithm design and model framework development;

• Developed an IoT-enabled intelligent irrigation solution for elevated greenery to optimize water usage;
• Implemented a Transformer-based model for accurate, on-demand water prediction;
• Used encoder-decoder architecture to enhance prediction accuracy;
• Trained predictive models with historical environmental data for real-world use;
• Designed a system that reduces workload and detects faults, lowering plant mortality;
• Led requirements gathering, algorithm research, and IoT sensor setup;

PUBLICATIONS & PATENTS

• Qijia Tian*, Yushi Feng*, Shuangyu Yang, Weifa Yang, Lequan Yu, MedDocBench: Benchmarking Vision-Language Models for Reliable Structured Extraction from Real-World Clinical Forms, Workshop on Structured Data for Health (SD4H) at ICML 2026 (Published).
• Yan Liu, Zhuyun Qi, Xingbo Feng, Shuangping Zhan, Jiashuo Lin, Chenxi Ling, Ruide Cao, Qijia Tian, Weichao Li, Yi Wang, ReCQF: Enhancing CQF Redundancy with Delay Alignment Scheduling in TSN, IEEE/IFIP Network Operations and Management Symposium 2026 (Accepted).
• Organic Solar Thin Film Cell Simulation Based on Transfer Matrix (Unpublished).

HONORS & AWARDS

Faculty of Engineering Peer Mentorship Programme Certificate 2023-2024	2024.07
Faculty of Engineering Peer Mentorship Programme Certificate 2022-2023	2023.07
Dean’s Honors List 2021-2022	2023.02
The 2022 RobotMaster Sim2Real Challenge First Prize	2022.08
Dean’s Honors List 2020-2021	2022.01

SKILLS & LANGUAGES

• Languages: Mandarin (Native), English (Work language), Cantonese (Basic)
• IT skills: Python, C, SQL, MATLAB, PyTorch, OpenCV, Vulkan SDK, LLM Prompt Engineering, Deep Learning, Verilog/VHDL

APPENDIX

The 2022 RobotMaster Sim2Real Challenge First Prize Team

Part of the results of the Final Year Project

Final Year Project Presentation

Senior Design Project Link

Certificates Copy

Medical AI Focus - English

Medical AI Focus - Chinese