We've talked about tuning, and we've talked about prompt engineering, but those are not the only techniques at our disposal to optimize LLMs.