About Wayne¶

Wei Cheng (Wayne) Chiu — Machine Learning Engineer.

I work on LLM inference and GPU performance — hand-written CUDA / Tensor-Core kernels, TensorRT-LLM and vLLM serving, and reproducible benchmarks on Hopper and Blackwell — alongside multi-agent and RAG systems taken end-to-end into production.

Most recently a Machine Learning Engineer at SYNCROBOTIC (through June 2026), and an M.S. in Computer Science from the National Taiwan University of Science and Technology (NTUST), researching AI/LLM security and privacy-preserving machine learning in Prof. Shao-Jui Wang's lab.

My recent work spans applied LLM systems (multi-agent orchestration, RAG, LLM evaluation) on DGX-class hardware, the NVIDIA inference and multi-GPU stack (TensorRT-LLM, Triton, NIM, NCCL, CUDA Tensor Cores), and privacy-preserving ML (federated learning, differential privacy, secure multi-party computation).

Focus areas¶

Deployment — taking AI systems into customer environments, from PoC through rollout and support.
Reliability — prompt/pipeline regression tests, output validation across quality and safety, distributed tracing.
Retrieval & agents — hybrid vector + keyword + knowledge-graph retrieval with semantic caching; planning / execution / validation orchestration, function calling and tool use.
Serving internals — KV-cache, quantization trade-offs, Flash Attention, CUDA kernels — enough to reason about cost and latency at design time.
Research — AI/LLM security & safety alignment; federated learning, differential privacy, secure multi-party computation.

Contact¶

Disclaimer

The views expressed on this site are my own and do not represent those of my employer or affiliated institutions.