Shivam Dwivedi@2k24csaiml1c241300987952

Edge-Pali is a revolutionary dynamic VRAM optimizer designed to bridge the gap between heavy generative AI models and edge device constraints. Standard vision-transformer models process images using a fixed 1024 patches, consuming significant memory with 512.0 KB per inference at 100% baseline accuracy. Edge-Pali introduces a learned-pruning architecture that dynamically reduces this to 213–239 patches based on scene complexity, achieving an impressive ~79% VRAM reduction while maintaining 97.15% accuracy. Unlike basic similarity heuristics that struggle to exceed 50% compression, our approach ensures production-viable efficiency for real-time applications. By implementing this within a containerized environment, Edge-Pali provides a seamless, scalable solution for edge deployment, proving that high-fidelity AI performance is achievable even on resource-limited hardware

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.