Back to portfolio

Case Study

RIMAS

AI Workflow System / In progress

Backend orchestration system for retail planning workflows, combining FastAPI, LangGraph, PostgreSQL persistence and audit events for human-reviewed recommendations.

RepositoryMore projects

Stack

FastAPILangGraphPostgreSQLMLflowDocker

Problem

Retail teams need repeatable intelligence workflows that can collect signals, coordinate agent steps and keep execution auditable as the system evolves.

Approach

The system is framed as a backend orchestration layer with FastAPI endpoints, LangGraph workflows, PostgreSQL persistence and MLflow tracking for experiment visibility.

Key Decisions

  • Use FastAPI as the operational API boundary.
  • Model the workflow as explicit orchestration instead of one-off scripts.
  • Keep persistence and experiment tracking visible through PostgreSQL and MLflow.

Tradeoffs

  • The project is still in progress, so the page focuses on architecture and direction rather than production metrics.
  • Multi-agent systems add coordination overhead, so auditability and constrained workflow design matter more than broad autonomy.