Chen-Wei (Wilson) Chang | Amazon SDE Intern and AI Researcher

About Me

I’m a Computer Science master’s student at Virginia Tech with experience building compiler systems, full-stack applications, and applied AI workflows. At Amazon, I designed a compiler that translates Agentic AI workflow DSLs into AWS Step Functions, and at TSMC I deployed Qwen3 on NVIDIA H100 GPUs to ship scalable inference APIs and automate anomaly analysis.

My research at Virginia Tech focuses on adversarial scam detection, model robustness, and hybrid systems that combine traditional ML with fine-tuned LLMs. I enjoy turning ideas into working systems and keeping research tied to deployment reality.

Experience

Software Development Engineer Intern

Amazon

June 2025 - August 2026

Designed and built a compiler that translates Agentic AI workflow DSLs into AWS Step Functions, replacing a legacy engine
Reduced workflow execution latency by 48% through compiler optimizations and native workflow state execution
Enabled cloud-native workflow orchestration with built-in durability, retry/catch, and execution observability
Verified behavioral parity across 1,200+ benchmark executions and production workflows through backend validation

AI Research and Development Intern

TSMC

June 2025 - August 2025

Deployed Qwen3 on H100 GPUs using SGLang, delivering scalable inference APIs for high-throughput applications
Built an Agentic AI system using OctoTools to automate FDC parsing, anomaly detection, and RAG-based SOP retrieval
Reduced manual data interpretation time by 70% by implementing logic-driven log analysis pipelines and visual reports
Improved anomaly resolution speed by 80% by optimizing incident response workflows, saving projected $620K annually

Research Assistant

Virginia Tech

August 2024 - May 2025

Reduced inference latency by 57% and optimized system performance by implementing selective escalation logic
Improved F1 to 0.90 and precision to 0.95 by adding a fine-tuned LLaMA tie-breaker to multi-model majority voting
Increased LLaMA-3 8B adversarial scam detection accuracy to 0.87 by applying LoRA with 4-bit quantization

Software Engineering Intern

Shin Kong Financial Holding

January 2023 - February 2023

Developed a credit card management system, integrating a Vue.js frontend with backend APIs to streamline workflows
Reduced data processing time by 90% by architecting Python automation scripts for large-scale data organization
Cut processing time by 50% for converting COBOL and DOT files to CSV with a Tkinter GUI

Education

M.S. in Computer Science

Virginia Tech, Alexandria, VA

08/2024 - 12/2026

Relevant Coursework: Software Engineering, Web Application Development, Database Management Systems, AI Tools for SWE

B.S. in Computer Science

National Dong Hwa University, Taiwan

09/2019 - 06/2023

Upper-Division GPA: 4.45 / 4.50

Relevant Coursework: Data Structures, Algorithms, OOP, Operating Systems, Computer Architecture

Hi, I'm Chen-Wei (Wilson) Chang