Overview

A major engineering project exploring natural-language access to student academic records. The system allowed teachers, department heads, and other authorized staff to ask questions about marksheets and student performance through a familiar chat interface instead of manually searching spreadsheets and PDFs.

The challenge

Academic records are often stored across structured files and documents, making questions such as who topped a class, how a student performed, or who ranked among the lowest scorers unnecessarily time-consuming to answer. Access also needed to respect the responsibilities and permissions of different academic roles.

Approach

I worked on a retrieval-augmented generation workflow that embedded and vectorized student-record documents so relevant information could be retrieved in response to natural-language queries. A conversational interface connected these retrieval capabilities to a role-based access-control system, allowing users to query records within their permitted scope.

Key features

• Natural-language queries over student academic records
• Retrieval-augmented generation pipeline
• Document embeddings and vector-based retrieval
• Question answering over student marksheets and records
• Role-based access control for academic departments
• Class topper and student-performance queries
• Cross-class queries for authorized department heads
• Conversational interface for academic-record discovery

Takeaway

The project explored how familiar conversational interfaces can make existing institutional data easier to access. It also highlighted that useful AI systems need more than retrieval and generation: authorization and the scope of each user's access are fundamental parts of the design.