Built a conversational agent that helps users learn and apply principles from an upcoming business and life management book.
Role
AI Engineer on a small team.
Challenge
The hard part was andragogy, the study of how adults learn. The agent could only teach well if it was built on what the research actually says works, not on intuition about good teaching.
Contributions
- Multi-Agent Architecture: Designed and built a hierarchical multi-agent architecture to handle complex conversational flows and teaching interactions
- Team Enablement: Trained the team on prompt evaluation and application observability tooling, and set the conventions they would use for agent development going forward
- Applied Learning Science: Ran a literature review on adult learning, then took what it surfaced and built it into how the agents taught, which improved how well users learned
Technologies
- Multi-agent orchestration
- Prompt engineering and evaluation frameworks
- Application observability tooling
- LLM-based conversational AI
