Experiences¶
Industry¶
SYNCROBOTIC — Machine Learning Engineer¶
Sep 2025 – Apr 2026 · New Taipei City, Taiwan · On-site
Sole developer of an enterprise multi-agent AI platform on DGX-class hardware — owned end-to-end from PoC through production, rolled out at two enterprise customers, partnering with each through deployment and troubleshooting.
- Scoped ambiguous enterprise requirements into a working multi-agent system, taken from PoC to production rollout.
- Unified local and commercial LLM backends under one routing layer, removing per-team model wrappers.
- Designed a planning–execution–validation orchestration with topological scheduling and parallel agent dispatch for multi-step workflows.
- Built a hybrid retrieval layer on LightRAG (vector + keyword + knowledge graph) with embedding-based semantic caching, cutting redundant LLM calls and keeping answers consistent on repeat queries.
- Owned production reliability — prompt/pipeline regression tests, output validation across quality and safety dimensions, distributed tracing; added CUDA-accelerated hot paths on critical inference.
Stack: Python · TypeScript · Rust · vLLM · LightRAG · NestJS · CUDA
Advantech — AI Agent Engineer¶
Jun 2025 – Aug 2025 · Taipei City, Taiwan · On-site
Built internal coding agents for developer tooling — function calling / tool use to automate code review, test generation, and static analysis across the SDLC (Claude Code-style assistants).
- Benchmarked tool-use reliability across commercial and local LLM backends.
- Designed reusable prompt templates and skill patterns adopted by the team.
TSMC (台積電) — Facility Engineer (R&D Alternative Military Service)¶
May 2021 – Oct 2022 · Tainan City, Taiwan
- Supported R&D fab facility operations at the world's leading semiconductor foundry under the national R&D alternative military service program.
Research¶
NTUST — Graduate Researcher, Prof. Shao-Jui Wang's Lab¶
Aug 2024 – Present · NTUST
- Research in AI/LLM security and safety alignment (defending against harmful fine-tuning and adversarial attacks) and privacy-enhancing ML (federated learning, differential privacy, secure multi-party computation).
- First-author of SelGrad: Selective Gradient Projection for Efficient Safety Alignment Against Harmful Fine-Tuning, under review at IEEE Transactions on Dependable and Secure Computing (TDSC).
- Differentially-private synthetic-data generation with MP-SPDZ (secure multi-party computation) for the NICS privacy-enhancing-technologies competition.
Education¶
National Taiwan University of Science and Technology (NTUST)¶
M.S. in Computer Science and Information Engineering · Aug 2024 – Apr 2026 · GPA 4.09
Research focus: LLMOps (fine-tuning, safety alignment), CUDA parallel programming, inference acceleration. Coursework: Machine Learning, Deep Learning, Parallel Programming, Cybersecurity.
National Taiwan University (NTU)¶
M.S. in Structural Engineering · Sep 2018 – Jun 2020
National Taiwan University (NTU)¶
B.S. in Civil Engineering · Sep 2014 – Jun 2018
Licenses & Certifications¶
- NVIDIA DLI — Fundamentals of Accelerated Computing with CUDA Python (Dec 2024)
- NVIDIA DLI — Fundamentals of Accelerated Computing with CUDA C/C++ (Nov 2024)
- (and 3 more — see LinkedIn)