Priority-Aware Model-Distributed Inference at Edge Networks

Short PA-MDI demo presented at IEEE ICMLCN 2025.

Communication-Aware Model-Distributed Inference via Latent Representation Compression

• I explored and implemented activation compressors and implemented, tested, and evaluated T5 (Flan-T5) and ResNet inference on NVIDIA Jetson across datasets. I also developed the project code in this repository.
• The repository includes accuracy-function fitting, Jetson online inference experiments, and performance trace collection.
• Experiments met throughput targets missed by uncompressed inference and achieved higher accuracy than baseline compression policies; Jetson tests also reduced inference delay at comparable accuracy.