Back to projects

AI-powered support ticket platform

SecureAI Desk / SentinelDesk

SecureAI Desk is a full-stack AI support-ticket system designed to improve manual ticket handling workflows. It uses AI-assisted classification and response support to help teams organize, prioritize, and respond to support tickets more efficiently.

SecureAI Desk / SentinelDesk project thumbnail

What I built

SecureAI Desk is a full-stack AI support-ticket system designed to improve manual ticket handling workflows. It uses AI-assisted classification and response support to help teams organize, prioritize, and respond to support tickets more efficiently.

Why I built it

Manual support-ticket triage is slow, inconsistent, and hard to scale when teams need reliable categorization, priority handling, and response support.

My role

Designed the backend API surface, ticket data model, authentication flow, AI classification workflow, and Dockerized development setup.

How it works

  • FastAPI service exposes authenticated ticket, user, and AI-assist endpoints.
  • PostgreSQL stores users, roles, ticket metadata, ticket history, and classification output.
  • AI service layer handles prompt construction, classification calls, and response suggestions.
  • Docker provides repeatable local setup for backend and database services.

System architecture

  • FastAPI service exposes authenticated ticket, user, and AI-assist endpoints.
  • PostgreSQL stores users, roles, ticket metadata, ticket history, and classification output.
  • AI service layer handles prompt construction, classification calls, and response suggestions.
  • Docker provides repeatable local setup for backend and database services.

Screenshots and demo

SecureAI Desk / SentinelDesk demo placeholder

Technical challenges

  • Designing AI output that stays useful and structured for downstream ticket workflows.
  • Balancing protected API design with fast iteration during prototype development.
  • Keeping sample metrics clearly marked until real production validation is available.

Results

  • 550 sample/support tickets processed
  • 90% classification accuracy placeholder, replace with validated result
  • 20+ protected backend endpoints

What I learned

  • Strong API contracts make AI workflows easier to test and reason about.
  • Support tools become more useful when automation assists humans instead of replacing context.

Explore the code

Explore the code, setup instructions, and technical documentation on GitHub.