PassDTS: AI-Powered Defense Travel System

First look

What I built

• Architected a full-stack platform with a React (Vite) frontend and a FastAPI (Python) backend.
• Built six cooperating AI agents: intake validation on Google Gemini, urgency classification, rule-based policy routing, an approval orchestrator state machine with SLA tracking, natural-language explanation and notification, and a background learning agent that analyzes denials and proposes routing fixes.
• Modeled users, requests, and approvals in PostgreSQL and containerized the services with Docker Compose behind a documented REST API.

What came of it

• Explainable AI reasoning surfaced for every routing decision in a real-time dashboard.
• A slow, manual authorization workflow modernized end to end.

Surfaces

Request Timeline

• Classified Mission Critical and routed to 3 approvers instead of the standard 5.
• Every step states the reason behind it, with the SLA clock and the remaining approval beside it.

Analytics Dashboard

• 1,247 requests at 2.3 days average processing.
• Volume trend, status distribution and cost trend across the fleet.

Create Request

• Seven sections, from basic info through justification and routing.
• The cost summary totals transportation, lodging and M&IE as the form is filled.