Meeting-to-Backlog Planner
Built a planning pipeline that turns meeting decisions into prioritized delivery stories, refining 350 proposals to 215 across ten transcripts.
Decisions into scoped delivery work · Role: Architect & Engineer
Architecture
- Transcripts are extracted concurrently, then reconciled in meeting order.
- Later decisions update earlier proposals instead of creating contradictions.
- Story writing runs in parallel; RICE scoring scopes the first delivery phase.
Results & scale
- 350 proposals reduced to 215 on the same ten-transcript input
- Approximately 13 minutes of end-to-end agent runtime
Technology stack
- LangGraph
- Python
- Azure AI Foundry
- ServiceNow
My contribution
Original extraction, chronological reconciliation, and backlog-generation pipeline.