AI People Wiki

Builders & open source

Xuan-Son Nguyen

Software engineer, Hugging Face

Xuan-Son Nguyen is a software engineer at Hugging Face and a core maintainer of llama.cpp, the C/C++ runtime for running language models on local hardware. The project's CODEOWNERS file lists Nguyen as the owner of its multimodal library and among the owners of llama-server, and Nguyen's site credits this work with adding vision and audio support to llama.cpp. The site also lists the Hugging Face Hub integration that lets Ollama run GGUF models hosted on the Hub, and Nguyen publishes wllama, a WebAssembly binding for running llama.cpp in the browser.1234

At a glance

Current role
Software engineer
Sector
Company
Work listed
3
Sources cited
12
Profile reviewed

Open-source work

Nguyen's llama.cpp contributions include work on its HTTP server, such as a January 2024 change that defers requests when no processing slot is free. The project's CODEOWNERS file now lists Nguyen as an owner of llama-server, shared common code, CI and Dockerfiles.62

In April 2025 Nguyen consolidated llama.cpp's separate vision command-line tools into a single multimodal tool, llama-mtmd-cli, merging the LLaVA, Gemma 3 and MiniCPM-V command-line programs. In May 2025 Nguyen added audio input, starting with the Ultravox model, and brought audio input to llama-server. CODEOWNERS lists Nguyen as the owner of the multimodal library, and Nguyen's site credits Nguyen with adding vision and audio support to llama.cpp.78921

Products and projects

At Hugging Face, Nguyen worked on the Hub's Ollama integration, which lets Ollama run public GGUF checkpoints from the Hub with a single command, and co-wrote the February 2026 announcement that the ggml and llama.cpp team was joining Hugging Face. Under Nguyen's own GitHub account, wllama provides a WebAssembly binding that runs llama.cpp inference in the browser.13104

Nguyen has written a personal blog since 2016. Recent machine learning posts include "Introducing: ggml-easy" (April 2025), a wrapper that simplifies working with GGML, an explainer on RoPE variants (April 2025), and a guide to using OCR models with llama.cpp (April 2026).1112

Work and contributions

  • llama.cpp

    Core maintainer, according to Nguyen's site, which also credits Nguyen with adding vision and audio support. CODEOWNERS names Nguyen for the multimodal library (llama-mtmd), llama-server, shared common code, CI, and Dockerfiles. Nguyen is also a byline author of Hugging Face's 2026 announcement that the ggml and llama.cpp team was joining the company.1210

  • Ollama integration on the Hugging Face Hub

    Listed among Nguyen's work. The Hub documentation describes running any public GGUF checkpoint on the Hub with a single ollama run command, with a choice of quantization type.13

  • wllama

    Published under Nguyen's GitHub account as an MIT-licensed WebAssembly binding for llama.cpp; it runs language model inference directly in a web browser.4

Timeline

  1. Opened a llama.cpp server pull request to defer tasks when no slot is available.6
  2. Introduced ggml-easy, a simple wrapper for working with GGML, on Nguyen's blog.12
  3. Had a pull request merged that combined the LLaVA, Gemma 3 and MiniCPM-V command-line tools into llama-mtmd-cli.7
  4. Added Ultravox audio input to llama.cpp's multimodal library.8
  5. Added audio input support to llama-server.9
  6. Co-wrote Hugging Face's announcement that ggml and llama.cpp were joining the company.10

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

12 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Xuan-Son Nguyen — about and workXuan-Son Nguyen · Official page · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  2. ggml-org/llama.cpp CODEOWNERSggml-org (GitHub repository) · Repository · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  3. Use Ollama with any GGUF Model on Hugging Face HubHugging Face · Documentation · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  4. ngxson/wllama READMEXuan-Son Nguyen (GitHub repository) · Repository · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  5. ngxson (Xuan-Son Nguyen) on GitHubGitHub · Official page · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  6. ggml-org/llama.cpp pull request #5018: Server: when no slot is available, defer the taskGitHub · Repository · published Jan 18, 2024 · accessed Oct 21, 2026
  7. ggml-org/llama.cpp pull request #13012: mtmd: merge llava, gemma3 and minicpmv CLI into single llama-mtmd-cliGitHub · Repository · published Apr 18, 2025 · accessed Oct 21, 2026
  8. ggml-org/llama.cpp pull request #13623: mtmd: add ultravox audio inputGitHub · Repository · published May 18, 2025 · accessed Oct 21, 2026
  9. ggml-org/llama.cpp pull request #13714: server: support audio inputGitHub · Repository · published May 22, 2025 · accessed Oct 21, 2026
  10. GGML and llama.cpp join HF to ensure the long-term progress of Local AIHugging Face · Announcement · published Feb 20, 2026 · accessed Oct 21, 2026 · reviewed Oct 21, 2026
  11. ngxson's blogXuan-Son Nguyen · Official page · accessed Oct 21, 2026
  12. Introducing: ggml-easy — ngxson's blogXuan-Son Nguyen · Announcement · published Apr 20, 2025 · accessed Oct 21, 2026