SZABOT

Overview

SZABOT is a retrieval-augmented academic assistant that parses timetable files and answers student questions in real time across classes, exams, and events.

The problem

Students and administrators often struggle to extract timely answers from scattered timetable data in Excel and PDF formats, especially when they need specific class, exam, or event information on demand.

The solution

I built a Python parsing pipeline to convert timetable files into structured data and combined it with Google Gemini-powered retrieval and intent detection logic. The assistant answers academic queries in real time while an admin dashboard supports schedule uploads and query monitoring.

Features

• Excel/PDF timetable parsing
• RAG-based assistant
• Student query answering
• Intent detection
• Admin dashboard
• Schedule upload workflows

Architecture

• Python ingestion layer for parsing Excel and PDF timetable documents into structured data
• RAG architecture combining retrieval with Gemini-powered reasoning for queries
• Intent detection routing questions across classes, exams, and events
• JavaScript-based admin dashboard for uploads and monitoring

Challenges

• Normalizing uneven timetable formats into reliable structured data for question answering
• Matching academic intents to the correct domain and return contextually correct answers

Results

The project delivers real-time academic assistance from structured timetable data, with an admin workflow to monitor queries and upload updates.