Exam Analytics System

Overview

Combines ML classifiers with agentic workflows to provide explainable difficulty analysis for exams.

Problem

Standard difficulty metrics lack "the why"—pedagogical explanations for why a question is hard.

Solution

A stateful LangGraph workflow that takes ML predictions and generates human-readable reasoning from an LLM.

Process

Trained scikit-learn models for base classification, then wrapped them in a LangChain agent pipeline.

Architecture

Python-based pipeline with a Streamlit interface and agentic reasoning at the core.

Tech Stack

Outcome & Results

An explainable AI system used for refining educational assessment tools.