Situation
AI automation becomes difficult to operate when long-running work, retries, approvals, and state live inside fragile synchronous requests.
Agentic automation
FlowPilot
A production-minded automation command center where AI workflows run as durable jobs instead of fragile synchronous prompts. Human approval, execution state, validation, and observability are first-class parts of the design.
Project story
Task
Build a production-minded AI automation control plane that treats agents as durable workflows instead of isolated prompts.
Action
- Separated API orchestration from execution with FastAPI, Redis/RQ workers, PostgreSQL persistence, and LangGraph workflows.
- Added human approval boundaries, evidence and validation for external-impact workflows, plus scheduled automation.
- Instrumented the system with Prometheus/Grafana metrics and OpenTelemetry/Jaeger traces and added CI evaluation coverage.
Result
- Created an operations-oriented architecture for content, competitor, outreach, and KPI automation patterns.
- Made workflow progress, audit state, approvals, retries, and observability first-class rather than hidden behind the LLM call.
Interactive demo · BYOK
Try FlowPilot live
Bring your own provider key to create a short-lived session for this project. The portfolio does not store the credential.
Used once to create the temporary portfolio session. It is not stored by the portfolio.
One session · reusable across project pages · provider key never stored