Conduct empirical experiments on parameter-efficient fine-tuning (LoRA/QLoRA), model quantization, and latency optimization on custom silicon.
Direct application reviewed by engineering leadership.