Ethan Tan@babyrico

Auto-clippers like ShadowPlay and Medal save everything — boring cleanup kills included. Salience is a personalized highlight-ranking layer that learns which moments YOU actually care about: one-pump shotgun kills, snipes, clutches, build fights, or Victory Royales. How it works: Salience watches the folder where clips are already saved, samples keyframes, and extracts structured labels (weapon, elimination type, context, victory). A ranked review inbox lets you mark clips favorite / keep / boring / skip, and a lightweight personal ranker updates instantly from that feedback plus an 18-question taste questionnaire. The AI story is teacher–student distillation. During development, Qwen VLM on Fireworks AI labels clips with 28 constrained highlight attributes and event timestamps. Those teacher labels train a local student: a MobileNetV3-small backbone with an event locator head, per-event weapon/evidence heads, and context heads, exported to ONNX and fused with local OCR (RapidOCR reads the weapon HUD and victory banner). On an 81-clip held-out eval, the student reaches ~0.80 mean per-label agreement with the VLM teacher — enemy-visible F1 0.97, elimination F1 0.93, victory F1 0.89 — while running entirely on CPU with no API key and no upload. Inference is sub-second per clip; the full pipeline including video decoding is ~30 s per clip. Privacy is the point: the shipped ranking loop never needs to upload gameplay. The cloud teacher is an optional development-time tool, and the same harness can route to AMD Developer Cloud (Qwen2.5-VL via vLLM) as the teacher endpoint.

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.