In July 2025, Ndavid contributed to the bytedance-iaas/vllm repository by implementing FP8 quantization and Gaudi inference support using Intel Neural Compressor. This work focused on optimizing machine learning model performance and efficiency for deployment on Intel Gaudi hardware. Leveraging Python and PyTorch, Ndavid connected the end-to-end workflow to enable accelerated model serving, reducing cost per inference and improving throughput. The integration of quantization techniques addressed hardware-specific requirements and established a foundation for future benchmarks and optimizations. The depth of this feature reflects a strong understanding of model optimization and quantization, though no major bug fixes were required during this period.