/
Tuning Copilot
Create your scoring system
Transform product specs, code, and user data into scoring systems that effectively tell good apart from bad.
Your quality engineering assistant
Here are your generated metrics
Analyzes sources
Extracts insights from usage data to determine what signals you're missing.
Brainstorms Signals
Determines what guardrails are needed for any goal you have in mind
Tunes scores
Tunes your final score based on stated requirements, and user feedback
Understand your requirements
Mine user data, PRDs, and system prompts for requirements that actually matter.
Score subjective/objective criteria
Generate natural language questions and python code to score subjective and objective criteria.
Understand your scorer
Explains scores and performance by generating tests and conducting experiments.
Tune to match user taste
Combine dimensions into a single score tuned from product requirements and user preferences.
Follow best practices
Ensure your dimensions perform effectively, together. Keeps them granular and independent.
Integrate instantly
Quickly get up and running by generating your scorer's integration code.
Build your scoring spec in seconds
Tell the copilot about your application, with PRDs, documentation, and system prompts.
Refine your scoring system by having your copilot create test cases, and edit the scorer for failing cases.
Integrate your scoring system into your code, or existing observability and eval tools.
Come back once you've collected more data, and customize your scoring even further.
