Solo builder exploring agentic infra + audio tooling on AMD MI300X for the hackathon's Track 3.
An intelligent multi-agent pipeline that uses vision-language models and a novel memory engine to automatically generate temporally-coherent 3D spatial audio for video.