MVP experiment
React · FastAPI · PostgreSQL
HR Management Platform
How can a support workflow organise incoming HR requests while keeping classification and response suggestions reviewable by people?
Frontend
Employee submits a demo request
The React and TypeScript interface collects the request and shows the submitted ticket in the employee demo view.
Backend
FastAPI classifies the ticket
FastAPI validates and stores the request. Deterministic mock logic returns a category, priority, suggested response and reasoning.
Review
HR or Admin reviews the suggestion
Demo review queues display the ticket and suggested response. A person remains responsible for deciding what to do.
What the MVP demonstrates
The project connects a React and TypeScript frontend to a FastAPI backend and PostgreSQL database. Its demo views support ticket submission, mock classification, suggested response drafts, filtering, sorting and ticket status updates.
What worked, what’s limited, and what I’d do next
What worked
Deterministic keyword rules make the mock classification predictable and runnable locally without API keys or external AI services. Returning the reasoning alongside the category, priority and draft makes the suggestion easier to review.
Current limitations
Keyword matching is limited and is not a real language model. The Employee, HR Team and Admin views are demos; real authentication and role-based access are not implemented.
What I’d do next
Expand backend tests for classification edge cases, add authentication and role-based access, and improve the review queues. Any future LLM integration would need validated structured output and a mock fallback.