I work at the intersection of computer vision, local LLM infrastructure, and practical AI tooling.
My research background includes collaborative work on image super-resolution, image restoration, stereo vision, and low-light enhancement, including papers in the NTIRE / CVPR Workshops community. Today, I focus on taking ideas beyond experiments: building systems that people can run, test, and use.
I enjoy self-hosted AI, developer tools, agent workflows, and open source.
| Project | What it does |
|---|---|
| ComfyUI-Prompt-Bookmarks | A lightweight ComfyUI sidebar extension for bookmarking, organizing, searching, and restoring reusable prompts without adding workflow nodes. |
| ComfyUI-MiVolo-V2 | ComfyUI nodes for MiVOLO v2 age and gender prediction. |
| VistaPair | A GPU-accelerated Rust image viewer for side-by-side comparison and visual QA. |
| PageVault | Self-hosted publishing for sharing AI-generated pages, Markdown, and other documents. |
| Codex Local Status Bar | A GNOME Shell extension that displays Codex quota information using locally recorded data. |
I have contributed as a co-author to collaborative computer-vision research, including:
- NTIRE 2018 Challenge on Single Image Super-Resolution: Methods and Results
- NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Results
- NTIRE 2023 Challenge on Stereo Image Super-Resolution: Methods and Results
- NTIRE 2024 Challenge on Low Light Image Enhancement: Methods and Results
These are selected examples, not a complete publication list. See my Google Scholar profile for the full list and current citation data.
- Local LLM systems: serving, routing, compatibility, and evaluation.
- AI agents: practical integrations, developer workflows, and tool interoperability.
- Computer vision: imaging workflows, restoration, and visual evaluation.
- Open-source software: small tools that solve real problems.
Research informs the ideas. Engineering makes them useful.