Worked on stabilizing continuous integration and expanding large-model build paths in the nod-ai/SHARK-Platform repository. Addressed CI instability for 70B-scale models by increasing the maximum iterations for the affinity solver in the model compilation script, enabling reliable single-device compilation for large models such as llama70B-fp16. This adjustment reduced flakiness in CI runs and accelerated feedback loops for large-model scenarios. Utilized Shell scripting to update model compilation workflows, focusing on performance tuning and model compilation. The work laid the foundation for scalable deployment of non-sharded, large-model workflows, ensuring that future development cycles for large models proceed with greater reliability and speed.