About

I'm an engineer working in applied machine learning, with a focus on multimodal models and search. Video Moment Finder is a personal project I build and maintain in the open — the whole codebase is public on GitHub under AGPLv3.

I started VMF after spending too many evenings scrubbing through long recordings — lectures, tutorials, gameplay captures — looking for one specific moment I half-remembered. Existing tools searched transcripts, not what was actually on screen. With multimodal embedding models like Qwen3-VL, that problem is finally tractable: you can describe a scene in plain language, or show an example image, and get the frames that look most like it.

VMF is intentionally small and single-purpose. If you have a bug report, a feature idea, or you just want to talk about multimodal retrieval, the best places to reach me are below.