Case StudyFull-Stack2026

University Research Knowledge Graph & Literature Intelligence

Nexus Scholar

An enterprise-grade academic research intelligence and automated knowledge graph platform inspired by the ORKG initiative. Ingests scientific literature (arXiv, DOIs, PDFs), extracts subject-predicate-object semantic triples, indexes 768d vector embeddings, and visualizes cross-disciplinary research connections in real time.

Live Demo

View Source Code

Key Highlights

PostgreSQL + pgvector · 768d Vectors · arXiv Ingestion · 2D/3D Force Graph

Automated arXiv & PDF Ingestion: Extracts RDF-like subject-predicate-object semantic triples instantly
N-Dimensional Matrix Comparisons: Side-by-side matrices contrasting model architectures and benchmark metrics
Interactive 2D/3D Knowledge Graph Topology: Real-time force-directed canvas mapping cross-department collaborations
Hybrid Semantic & Vector Search: 768d dense vector indexing over research literature with pgvector

Tech Stack & Tools Used

React 18

TypeScript

Vite

Serverless API

PostgreSQL

pgvector

Prisma

Python

FastAPI

01. Problem

Academic literature is trapped in isolated PDF silos, requiring researchers to spend dozens of hours manually comparing benchmark metrics and architectures.

02. Design

Architected a dual-view portal combining high-density matrix comparison tables with an interactive 2D/3D force-directed knowledge graph.

03. Build

Engineered high-throughput arXiv ingestion microservice, PostgreSQL with pgvector indexing, and lightweight React/Vite visualization frontends.

04. Outcome

Deployed zero-cold-start research exploration platform accelerating literature reviews by up to 90%.

Outcome

Cut literature review duration by 90% via automated semantic triple extraction and interactive comparative matrix generation.

Demonstrates end-to-end software execution, clean UI polish, and practical problem solving.

Back to Portfolio

MRG.

Initializing Experience...

Back to portfolio