mb.
Forward-Deployed AI Engineer @ Adobe
← Back to all AI projects
PERSONAL PROJECT / DOCUMENT INTELLIGENCE & RAG

EduTask

From a dense academic syllabus to an actionable conversation.

LangChain Retrieval-Augmented Generation (RAG) Vector Database Python Flask Unstructured API

The Problem

Academic syllabi contain crucial deadlines, grade weightings, exam dates, and course policies scattered across 15+ page documents. Students frequently miss submission dates or misunderstand policy nuances simply because relevant context is buried in prose.

EduTask was built to bridge this disconnect: providing an intelligent retrieval system that ingests unstructured document formats (PDFs, DOCXs) and enables users to ask precise questions like "What is the late submission policy?" or "Extract all homework deadlines into a list."

System Architecture

The system combines the Unstructured API for document element extraction, LangChain for vector orchestration and contextual retrieval, and a vector database for semantic chunk retrieval.

01 Document Ingestion (Syllabus PDF/DOCX)
02 Parsing & Chunking via Unstructured API
03 Vector Embedding & Similarity Index
04 Context-Augmented LLM Response Generation

Recorded Demo

Recorded application walkthrough demonstrating document ingestion, semantic query answering, and automated task extraction.

Next Steps & Navigation

This experiment proved how domain-specific RAG pipelines can turn passive documentation into active, conversational knowledge bases.

Next project: Anime Recommender System →