Personal projects · Local AI + Multi-Agent Systems
The problem
Two workflows I wanted to speed up: talking to an AI assistant without sending everything to the cloud, and turning software requirements into design artifacts (UML) without manual diagramming.
Approach
- Voice agent — a locally-hosted real-time voice agent built on a local speech recognition pipeline
- UML generator — a multi-agent system that converts software requirements into UML outputs
Stack
- Local speech recognition pipeline
- LLM-based multi-agent orchestration
Outcome
Improved rapid ideation and requirement-to-design translation — keeping sensitive conversations local while automating the boring part of design docs.
[ screenshots + repo link coming soon ]