Phoenix - Your AI AssistantSystem Prompt Monitoring — Track AI Behaviour Drift
Delivery in
5 days
- Views 12
Amount of days required to complete work for this Offer as set by the freelancer.
Rating of the Offer as calculated from other buyers' reviews.
Average time for the freelancer to first reply on the workstream after purchase or contact on this Offer.
What you get with this Offer
I will build a system prompt behaviour monitoring system — tracking whether your AI's actual production behaviour remains consistent with your system prompt's intended behaviour, detecting when model updates, fine-tuning changes, or prompt injection attacks cause behaviour drift, and alerting when significant deviations are detected. System prompt behaviour monitoring addresses a problem that most AI operators don't discover until it causes a user incident — LLM providers update their models regularly, and a model update can change how the same system prompt is interpreted, causing behaviour changes that are invisible without systematic monitoring of AI output characteristics.
The system covers AI response sampling, behaviour characteristic extraction (tone, constraint adherence, format compliance, topic boundary adherence), drift detection against a baseline, alert configuration for significant deviations, and a behaviour dashboard showing trends over time.
The system covers AI response sampling, behaviour characteristic extraction (tone, constraint adherence, format compliance, topic boundary adherence), drift detection against a baseline, alert configuration for significant deviations, and a behaviour dashboard showing trends over time.
What the Freelancer needs to start the work
Please share your system prompt, your behaviour characteristics for monitoring, your LLM provider, your sampling rate, your alert channel, and your baseline period for drift comparison.