"A Computing Task Allocation System and Method for Distributed Inference in Large Language Models." This patent focuses on optimizing the inference process of LLMs on mobile devices by utilizing distributed task allocation, model segmentation, and linear optimization techniques. This patent is based on the February 2024 prototype and experimental results from the DigiMorph Lab, demonstrating real-world feasibility and performance improvements in decentralized AI inference. Feature Google DeepMind (Edge AI) NVIDIA Megatron & DistilBERT Huawei & Qualcomm Edge AI This Patent (DigiMorph) Model Optimization Quantization-based LLM inference Model parallelism & knowledge distillation AI inference acceleration for mobile devices Distributed task allocation across multiple devices Computational Approach Single-device execution High-performance clusters Edge-device AI acceleration Mixed-Integer Linear Programming (MILP) for dynamic task distribution Communication Handling Assumes stable network No focus on network efficiency Optimized for proprietary AI hardware Models packet loss, jitter, and bandwidth constraints Scalability Limited by device constraints Requires dedicated AI hardware Optimized for proprietary mobile AI chips Scales across edge devices with dynamic scheduling Energy Efficiency Moderate High-performance but energy-intensive Optimized for Qualcomm/Huawei AI processors Uses free/low-cost local compute with efficient task scheduling Know More DigiMorph Lab