NexusQ is an enterprise-grade academic intelligence platform designed to move beyond passive learning management systems. By constructing a localized "Digital Twin" of university curricula, NexusQ applies GraphRAG (Graph Retrieval-Augmented Generation) and multi-agent consensus to quantify knowledge gaps, predict assessment trends, and optimize learning pathways for both students and institutions.
The platform operates on a four-layer neuro-symbolic stack designed for high-fidelity reasoning and minimal hallucination.
- Multimodal Decomposition: Utilizes Vision Transformers (ViT) and GPT-4o Vision to extract semantic meaning from complex diagrams, equations, and handwritten notes.
- Atomic Chunking: Deconstructs course material into fundamental "knowledge atoms," enabling precise retrieval and relationship mapping.
- Privacy-Preserving Federation: Sanitizes sensitive data locally before aggregation, ensuring compliance with institutional privacy standards.
- GraphRAG Engine: powered by Neo4j, mapping semantic relationships between concepts (e.g., "Entropy" → "Second Law" → "Carnot Cycle").
- Cross-Institutional Super-Nodes: Identifies universal academic constants across disparate university syllabi.
- Temporal Drift Analysis: Quantifies how curriculum focus shifts over 5-10 year horizons using time-series vector embeddings.
A "Debate Architecture" where distinct AI agents argue to reach a consensus probability:
- The Historian: Analyzes longitudinal frequency of topics in past assessments.
- The Profiler: Evaluates instructor-specific patterns, biases, and areas of emphasis.
- The Auditor: Detects outliers and "anti-patterns" (topics overdue for assessment).
- The Adjudicator: Synthesizes conflicting signals into a final, calibrated probability score.
- Visual Confidence Heatmaps: Real-time rendering of student proficiency against the knowledge graph.
- Dynamic Problem Generation: Creating isomorphic variations of predicted problems for robustness training.
- Low-Bandwidth Optimization: Engineered for high performance on high-latency networks (2G/3G/Edge).
Refer to INSTRUCTIONS.md for detailed deployment guides.
| Tier | Target Audience | Key Capabilities |
|---|---|---|
| Standard | Individual Students | Basic Trend Analysis, OCR Ingestion, Community Access |
| Scholar | Power Users | GraphRAG Access, Multi-Agent Predictions, LaTeX Export, Unlimited History |
| Industrial | Universities | Departmental Analytics, Curriculum Alignment, White-Label Reports, SSO |
Frontend Infrastructure
- Framework: Next.js 14 (App Router)
- Language: TypeScript 5.0
- UI System: Tailwind CSS, Radix UI
- State Management: React Query, Zustand
Backend Services
- Orchestrator: Python 3.12 (FastAPI)
- Reasoning: LangGraph (Multi-Agent Workflows)
- Vector Search: Qdrant / ChromaDB
- Symbolic Logic: SymPy (Neuro-symbolic verification)
Data & DevOps
- Graph Database: Neo4j
- Relational Database: PostgreSQL (Prisma ORM)
- Containerization: Docker & Docker Compose
- CI/CD: Vercel / GitHub Actions
- Node.js 18+
- Python 3.12+
- Docker Desktop (for local Graph/Vector DB)
-
Clone Repository
git clone https://github.com/SourishSenapati/nexusq-academic.git cd nexusq-academic -
Initialize Frontend
cd frontend npm install # Config environment variables in .env npx prisma generate npm run dev
-
Initialize Backend
cd ../backend pip install -r requirements.txt python -m uvicorn main:app --reload -
Launch Infrastructure
docker-compose up -d
- Q1 2026: Federated Learning (Privacy-preserved multi-org resource sharing).
- Q2 2026: Mobile App (React Native) for offline access.
- Q3 2026: Regional Language Support (Hindi, Bengali, Tamil).
- Q4 2026: AI Tutor (Chat with Exam Paper) and Institutional API integration.
We welcome contributions from the academic and open-source community. Please review CONTRIBUTING.md for our code of conduct and pull request standards.
This project is licensed under the Apache 2.0 License - see the LICENSE file for details.
NexusQ Systems • Kolkata, India