Skip to content
View wu840407's full-sized avatar
💭
AI infra & speech-AI engineer · self-hosted LLM · ASR · PyTorch/CUDA
💭
AI infra & speech-AI engineer · self-hosted LLM · ASR · PyTorch/CUDA

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Block or report wu840407

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
wu840407/README.md

Hi, I'm Roger (ChengRung Wu) 👋

AI infrastructure & speech-AI engineer. 9 years building and deploying mission-critical systems — now focused on self-hosted LLM infrastructure and low-resource speech AI.

🚀 Featured Projects

Project What it is
enterprise-airgapped-llm Production self-hosted LLM platform for air-gapped, security-critical environments — 30B MoE across dual GPUs, fully offline.
YaYan-AI Offline multi-dialect speech-intelligence system — 22 Chinese dialects + 40 languages, speaker diarization.
openclaw-agents Autonomous LLM agents — OODA decision support & news-intel briefing.

🛠 Tech Stack

Python · C++ · PyTorch · CUDA · vLLM · Docker · Kafka · Linux · Elasticsearch · AWS

🎓 Background

M.S. CS @ NYCU · AWS Solutions Architect · CEH · HITCON white-hat · CVE research

📫 Contact

wu840407@gmail.com — open to AI-infra / FDSE / backend roles (2027)

GitHub stats

Pinned Loading

  1. YaYan-AI YaYan-AI Public

    Fully-offline multi-dialect speech-intelligence system — 22 Chinese dialects + 40 languages, speaker diarization, LLM-assisted correction (Whisper + Qwen + pyannote).

    Python 1

  2. enterprise-airgapped-llm enterprise-airgapped-llm Public

    Production reference architecture for self-hosted LLM in air-gapped enterprise environments. Dell R740 + Turing GPUs + vLLM + Qwen3-Coder + AD/LDAPS.

    Shell

  3. openclaw-agents openclaw-agents Public

    Four standalone OpenClaw agents — Gwen (OODA strategy receptionist), Miku (dual-MA futures trader), Jolly (multi-platform social poster), Lay (T-1 news briefing aggregator). Persona-driven, Telegra…

    Python

  4. MultiBodyCuboids MultiBodyCuboids Public

    M.S. thesis — multi-body motion-segmentation for arbitrary numbers of disordered point sets (PyTorch).

    Python

  5. yolov5-parallel yolov5-parallel Public

    Multi-GPU parallelized YOLOv5 training / inference pipeline.

    Python 2

  6. WIRE5.0 WIRE5.0 Public

    VR weapons-interactive training & virtual-museum system (HTC VIVE + Leap Motion).

    C#