System notes
Provue Multi-Agent Platform
Pazago
Built the backend and RAG foundation of a configurable AI agent platform, from ingestion and retrieval through tools and model configuration.
Context
I joined Provue as its first engineer and built the initial backend and RAG foundation. The product later grew to more than 10,000 registered users and approximately 100 production agents.
What I built
- Agent creation and configuration workflows, prompt and instruction management, MCP tools, plugins, and configurable LLM support.
- A Mastra-based RAG pipeline using pgvector for knowledge retrieval.
- A reusable user-level knowledge library and tone profiles that could be shared across agents.
- AWS infrastructure and CI/CD using ECS, ECR, RDS, Cognito, S3, CodePipeline, and Amplify.
Leadership
As the product scaled, the team expanded to eight backend and frontend engineers. I led the team while remaining hands-on with architecture and core backend development.