Whenever you're ready, walk me through how you'd approach this.
How do you ensure data security and privacy when fine-tuning or prompting models?
Explain a practical machine learning workflow covering data collection, preprocessing, storage, training, inference, monitoring, and incident response. Address personally identifiable information, confidential business data, access control, encryption, retention, logging, third-party model providers, prompt injection, memorization, and output leakage. Include how you would validate that privacy controls do not materially damage model utility, and provide production-quality Python code for detecting and redacting sensitive content before training or inference.