His research is centered on the cutting-edge domain of Large Language Models (LLMs). His work explores the intricate mechanisms behind these powerful AI systems, investigating aspects such as model architecture, training methodologies, and ethical considerations in AI deployment. Passionate about artificial intelligence, he is actively engaged in AI-related projects. He also shares his knowledge by participating in technical seminars and workshops, and by writing articles on his personal blog. Through these activities, he aims to foster a broader understanding and appreciation of AI and LLMs.
Hi, I’m Wei-Lin Wen (溫為霖) — an AI Engineer & Researcher from Taiwan 🇹🇼, specializing in Large Language Models (LLMs), NLP systems, and end-to-end generative AI integration.
Currently a Master’s student at National Yang Ming Chiao Tung University (NYCU), I build AI systems that bridge cutting-edge research and real-world deployment.
🧑💻 Who Am I
I am a Computer Science Master’s student at National Yang Ming Chiao Tung University (NYCU), passionately exploring the frontiers of Natural Language Processing (NLP) and Large Language Models (LLMs). My academic journey is marked by excellence, having graduated in the top 1% of my undergraduate class and being inducted as a Phi Tau Phi Honorary Member.
My technical core lies in end-to-end AI system integration. I specialize in bridging the gap between cutting-edge AI research and practical, deployable software solutions. I have successfully engineered multiple generative AI applications, leveraging techniques such as Supervised Fine-Tuning (SFT) and Retrieval-Augmented Generation (RAG) to automate complex workflow tasks.
Key Technical Strengths & Experiences:
✨ Generative AI & System Integration: I have extensive experience building agentic workflows and automated summarization systems. My work involves fine-tuning open-source models and integrating them into robust backend systems to deliver high-quality, readable outputs for complex data processing tasks.
🧑💻 High-Performance Computing (HPC): Beyond model architecture, I am proficient in managing the infrastructure that powers AI. I developed a “vLLM Manager” to dynamically orchestrate vLLM servers across dual DGX Spark clusters, optimizing resource allocation and system throughput. My A+ performance in Parallel Programming (N-body Simulation with 3D Ray Tracing) demonstrates my ability to write highly efficient, hardware-aware code.
✨ Advanced RAG Implementation: I explored Agentic RAG frameworks to analyze semiconductor industry earnings calls, showcasing my ability to build systems that can reason through specific domain knowledge and retrieve precise information.
🔬 Research & Optimization: First author of a paper presented at TANET 2024 regarding LLM optimization benchmarks on GPUs, reflecting my commitment to understanding and improving model inference efficiency.
Awards & Achievements:
🎖️ 1st Place, International ICT Innovative Services Awards 2024 (AI Tools Category) – Top 1 out of 1623 teams.
🎖️ 1st Place, Smart Sustainable Circular Technology Workshop & Project Competition (2024).
🎖️ Phi Tau Phi Honorary Member – Recognition for top-tier academic performance.
📄 Research Bio (Journal Publication)
🚀 Core Technical Competencies
- LLM Fine-Tuning: Supervised Fine-Tuning (SFT) on open-source LLMs (LLaMA, Qwen, Gemma, etc.)
- RAG Systems: Agentic RAG, multi-hop retrieval, domain-specific knowledge integration
- Generative AI: Prompt engineering, LLM-as-a-Judge, automated evaluation pipelines
- Deep Learning: UNet, ResNet, GAN (StyleGAN-NADA), Binary Semantic Segmentation
- NLP: Text Summarization, Clinical Report Generation, Named Entity Recognition
- HPC / GPU Cluster: vLLM server orchestration across Dual DGX Spark Clusters
- Parallel Computing: CUDA, MPI, OpenMP — A+ in Parallel Programming (Rank 3/130)
- Cloud: AWS (S3, EC2), Google Cloud
- DevOps: Docker, Linux sysadmin, network administration, TrueNAS (RAID/SMB/iSCSI)
- Languages: Python · C++ · C · Java · JavaScript · TypeScript
- Backend: Node.js, RESTful API design, FastAPI
- Frontend: Vue.js, HTML5, CSS3
- Tools: Git, Google Analytics, Google OAuth
🏆 Highlights & Awards
🥇 1st Place — International ICT Innovative Services Awards 2024, AI Tools / Generative AI Category (Top 1 of 1,623 teams, ~0.06%)
🥇 1st Place — Smart Sustainable Circular Technology Workshop & Project Competition 2024 (College Group, Top 1 of 26 teams)
🎓 Phi Tau Phi Honorary Member — Awarded to the top 1% of graduates nationally in Taiwan
🎓 Academic & Conduct Excellence Graduate Award — Highest combined academic + conduct score in the graduating class
📜 First Author, TANET 2024 (Oral Presentation) — “Benchmarking Optimization Techniques for Large Language Models on GPUs” — Wei-Lin Wen, Yu-Yao Tsai, Chao-Tung Yang
📺 TVBS News Interview — Featured for generative AI capstone research project (July 2024)
💼 Experience
🏫 Teaching Assistant · NYCU · 2026 – Present
Natural Language Processing — National Yang Ming Chiao Tung University
- Designed and delivered lab curriculum on LLM-as-a-Judge evaluation methodology
- Guided graduate students through NLP concepts, model evaluation, and hands-on assignments
🔬 Research Assistant · Tunghai University (NSTC Project) · 2025 – 2026
Project: SFT LLMs for In-Hospital Nursing Record Summarization
- Fine-tuned open-source LLMs to automate consolidation of multi-shift nursing records
- Engineered an end-to-end pipeline: raw clinical text → structured, human-readable summaries
- Delivered measurable improvements in documentation efficiency for clinical nursing staff
🏥 Research Assistant · Taichung Veterans General Hospital · Feb – Jul 2025
(1) Emergency Department — Rong-Dong Project
- Developed a Generative AI System for Emergency Discharge Summary Summarization
- Automated extraction and synthesis of multi-source patient records into concise discharge reports
(2) Cardiology Department
- Collaborated on a High-Readability Echocardiography Report Generation System
- Applied SFT on LLMs to produce accurate, physician-ready cardiac ultrasound reports from raw data
🎓 Education
🏛️ M.S. Computer Science & Engineering · NYCU · Sep 2025 – 2027 (Expected)
National Yang Ming Chiao Tung University (國立陽明交通大學 - 資訊科學與工程研究所 - 甲組)
Institute of Computer Science and Engineering
Research Focus: Deep Learning · Natural Language Processing · Large Language Models
Lab Role — Lab Network Administrator (網管)
Portfolio: vLLM Manager for Dual DGX Spark Cluster
Designed and developed a dynamic orchestration system for managing vLLM inference servers across two DGX Spark GPU nodes. Enables concurrent multi-user inference with automatic resource allocation and load balancing.
Key Coursework:
| Course | Grade | Score | Rank |
|---|---|---|---|
| Parallel Programming (平行程式設計) | A+ | 97.14 | 3 / 130 |
| Generative Information Retrieval (資訊檢索與擷取) | A+ | 93.22 | 2 / 80 |
| Data Mining (資料探勘) | A+ | N/A (Grade Only) | N/A (Grade Only) |
- Final Project (Parallel): N-body Simulation with 3D Ray Tracing — Parallelization Strategies & Performance Analysis
- Final Project (Gen. IR): Earnings Call Analysis via Agentic RAG in the Semiconductor Industry
🏛️ B.S. Computer Science · Tunghai University · Sep 2021 – Jun 2025
Tunghai University
Department of Computer Science and Information Engineering
Class Rank: 1 / 61 (Top 1.64%) · Overall Grade: A+
Honors: Phi Tau Phi Honorary Member · Academic & Conduct Excellence Graduate Award · Lifetime Member, College of Engineering Honor Society · University Highlight Program Scholarship: NT$240,000 (~USD 8,000)
Competitions & Publications:
- 🥇
1st Place— International ICT Innovative Services Awards 2024 (AI Tools, Generative AI Category) — #1 of 1,623 teams - 🥇
1st Place— Smart Sustainable Circular Technology Competition (College Group) — #1 of 26 teams - 📄
First Author, TANET 2024 (Oral): “Benchmarking Optimization Techniques for Large Language Models on GPUs”
Industry-Academia Projects (產學合作計畫):
NSTC Project (國科會計畫)— Nursing Record Consolidation Assistant SystemKuang Tien General Hospital (光田綜合醫院)— PrivNurse AI (Local SFT-LLM nurse assistant)Taichung Veterans General Hospital (臺中榮民總醫院)— Emergency & Cardiology AI report generation systems
Key Coursework:
| Course Title | Grade | Course Title | Grade |
|---|---|---|---|
| Data Structures (資料結構) | A+ | Operating Systems (作業系統) | A+ |
| Algorithms (演算法) | A+ | Compilers (編譯器) | A+ (99) |
| Database Systems (資料庫) | A+ | Systems Programming (系統程式) | A+ (99) |
| Logic Design (邏輯設計) | A+ | Computer Organization (計算機組織) | A+ (99) |
| Computer Networks (計算機網路) | A+ (99) | Introduction to Computer Science (計算機概論) | A+ |
| Object-Oriented Analysis and Design (物件導向分析與設計) | A+ | Cloud Computing (雲端計算) | A+ |
🔭 Selected Projects
🖥️ vLLM Manager — Dual DGX Spark Cluster Orchestration
Built an internal infrastructure tool for the lab that dynamically manages vLLM inference servers across two DGX Spark nodes. Supports concurrent multi-researcher usage with intelligent GPU resource scheduling, health monitoring, and zero-downtime server restarts. Significantly reduced idle GPU time and inference bottlenecks across the team.
Stack: Python · vLLM · NVIDIA DGX · REST API · Linux
🏥 PrivNurse AI — Privacy-First Local LLM Nurse Assistant
Designed and deployed a fully on-premise AI nurse assistant system for Kuang Tien General Hospital. Fine-tuned open-source LLMs via SFT on proprietary nursing datasets to deliver accurate clinical decision support — with zero patient data leaving the hospital network.
Stack: Python · SFT (LLaMA/Qwen) · FastAPI · Local GPU Inference
📈 Agentic RAG for Semiconductor Earnings Call Analysis
Engineered an agentic Retrieval-Augmented Generation system that ingests and reasons over semiconductor industry earnings call transcripts. Supports multi-hop reasoning, financial signal extraction, and domain-specific Q&A with verifiable source attribution.
Stack: Python · LangChain / LlamaIndex · Vector DB · LLM API
🏥 Emergency Discharge Summary System (Rong-Dong Project)
Deployed at Taichung Veterans General Hospital, this system automates the generation of structured discharge summaries from multi-source emergency patient records using generative AI. Reduced manual documentation time for ED physicians and improved summary consistency and completeness.
Stack: Python · LLM (SFT) · Clinical NLP · FastAPI
🔬 Binary Semantic Segmentation (UNet & ResNet34-UNet)
Implemented, trained, and benchmarked UNet and ResNet34-UNet architectures from scratch for binary semantic segmentation. Conducted systematic ablation studies across encoder depth, skip-connection strategies, and data augmentation pipelines.
Stack: Python · PyTorch · UNet · ResNet34
🎨 StyleGAN-NADA — Text-Driven Domain Adaptation GAN
Implemented and fine-tuned a StyleGAN-NADA model for zero-shot domain adaptation of generative image models using CLIP-guided optimization. Explored how text prompts can steer a pre-trained generator into unseen visual domains without paired training data.
Stack: Python · PyTorch · StyleGAN · CLIP
📦 More Projects
- C Compiler Implementation — Full compiler pipeline: lexer → parser → intermediate code generation
- Movie Recommender with Fuzzy Logic — Rule-based recommendation engine using fuzzy inference
- Online Image Hosting Platform — Full-stack cloud app with AWS S3 + Google OAuth
- University Lecture Hall Booking System — Web scheduling system with admin dashboard
- ChatGPT Telegram Bot — Conversational AI assistant integrated into Telegram via OpenAI API
- TrueNAS Lab Server Setup — Configured RAID arrays, SMB shares, and iSCSI targets for lab NAS
- Digital Twin for Auto-Irrigation System — IoT sensor integration + digital twin simulation
- Construction Safety Equipment Detection — Computer vision model for PPE compliance detection
- 3-Player 4×4 Tic-Tac-Toe in SIC Assembly — Low-level game implementation in assembly language
📜 Certifications
🟢 NVIDIA® Deep Learning Institute — Getting Started with AI on Jetson Nano
🟢 NVIDIA® Deep Learning Institute — Disaster Risk Monitoring Using Satellite Imagery
🟢 Google Analytics Certification
🌐 Connect with Me
I’m currently open to R&D, AI Engineering, or Software Engineering roles / internships — particularly those involving LLM system design, Agentic AI system, or GPU infrastructure.
Feel free to reach out! 🤝
| weilinwen1205@gmail.com | |
| linkedin.com/in/weilin-wen | |
| 🐙 GitHub | github.com/weilin1205 |
| 🌐 Blog | https://weilin1205.github.io |
“The universe is uncertain — you and I are dark horses. 🐎”
