ai researcher in metalearning, prompt optimization, evolutionary learning, RL

# ai researcher in metalearning 

- client: ai product development company from germany
- product: cogit - ai agent-based automation platform
- experience: machine learning, self-optimization systems, prompt engineering, llm evaluation
- tasks:
   - architect the agent.maker meta-framework with sub-agents (instruction.maker, config.maker, screen.maker)
   - develop self-improvement loops for iterative prompt refinement
   - build evaluation framework for agent output quality (instructions.md, config.yaml generation)
   - implement llm observability and tracing (phoenix/arize integration)
   - create quality metrics for generated agents (tool usage accuracy, task completion rate)
   - develop prompt mutation strategies based on execution feedback
   - build fitness evaluation for config parameters (model selection, temperature tuning)
   - implement learning mechanisms from agent execution logs
   - create benchmarks comparing models (gemini vs claude) for different agent tasks
- languages: python
- libraries: google adk, anthropic sdk, phoenix tracing, pydantic, yaml