Projects

Projects

Research & Development Projects

A comprehensive portfolio of research, applied systems, and future concepts bridging construction engineering and artificial intelligence.

10+ Total Projects
4 Published Systems
82.5% NL Accuracy
125+ Research Citations

🔬Doctoral Research Projects

Core PhD research projects developing novel frameworks and systems for construction knowledge management.

Web Application
Published

IproK Web Platform for Project Data Management

Developed a comprehensive web-based application serving as the user interface for the IproK ontology. Enables non-technical project managers to easily manage, visualize, and interact with integrated construction data in real-time.

Key Features

  • Project creation and planning workflows
  • Real-time dashboard visualization with NetworkX
  • Interactive scheduling and resource management
  • Ontology manipulation with Owlready2
Streamlit Owlready2 NetworkX Python RDF/OWL
NLP System
Published

ChatGraphDB: LLM-Powered Conversational Interface

Developed a conversational AI interface for querying project knowledge graphs using natural language. Built a dual-web-application framework using Streamlit and fine-tuned GPT-4 with ontology-aware prompts to translate questions into SPARQL queries.

Performance Metrics

82.5% Translation Accuracy
~5s Avg Response Time
100% IFC Coverage
📹 ChatGraphDB Demo & Query Examples
GPT-4 SPARQL Streamlit RDFlib IFC/BIM
Machine Learning
Published

Predictive Analytics Using Graph Neural Networks

Created a predictive model to forecast task-level delay and cost-overrun risks in construction projects. Engineered temporal, resource, and cost features from the IproK knowledge graph and used this data to train Graph Neural Network models.

Key Contributions

  • Feature engineering from knowledge graphs
  • Temporal dependency modeling
  • Risk pattern identification
  • Proactive project control system
PyTorch Geometric DGL Graph Neural Networks Python Knowledge Graphs
Integration Framework
Published

Ontology-Driven Bi-Directional BIM/IFC Workflow

Architected a novel bi-directional workflow to integrate project management data (from the IproK ontology) with building data (from the IFC standard). Developed Python-based process to programmatically generate new, enriched IFC models from the unified knowledge graph.

Key Achievements

  • Registered ontology at w3id.org/iprok/
  • Seamless BIM-to-ontology conversion
  • Ontology-to-IFC generation
  • Standards-compliant data consistency
Python IFC ifcopenshell RDF/OWL BIM

⚙️Applied & Personal Projects

Practical applications and tools demonstrating real-world implementation of AI and BIM technologies.

Document Intelligence
Ongoing

Multimodal Agentic RAG for Construction Contracts

Developing an agentic Retrieval-Augmented Generation system for querying multimodal construction documents. The system architecture combines Docling, an ontology, Graphiti, a Vector DB, and a custom AgenticRAG flow.

Capabilities

  • Text, diagram, and table processing
  • Context-aware information extraction
  • Accurate retrieval from complex documents
  • Support for contracts, tenders, reports
📹 Multimodal RAG System Walkthrough
Docling LangGraph ChromaDB Graphiti RAG
Smart City AI
Ongoing

A.U.R.A.: AI Urban Reporting Agent for Smart Cities

Architected "A.U.R.A." (Amaravati Urban Resilience Agent), a collaborative multi-agent AI system for smart city data. Utilized IBM watsonx Orchestrate for agent collaboration and developed a custom data pipeline using QGIS and a new ontology (ADTO) to convert geospatial data into an RDF knowledge graph.

System Features

  • Multi-agent collaboration framework
  • Custom ADTO (Amaravati Digital Twin Ontology)
  • Geospatial data to RDF conversion
  • Intelligent urban infrastructure querying
IBM watsonx QGIS RDF Multi-Agent Systems Geospatial

🔮Future Research Concepts

Visionary projects exploring the next generation of construction intelligence and digital twins.

Concept
Future Work

Digital Twin of Amaravathi with A.U.R.A. Integration

A large-scale, high-fidelity digital twin of the city of Amaravathi. Planned for development using Unreal Engine for physics-based simulation, Cesium for lightweight 3D tiles, and an AI Assistant for conversational interaction.

Proposed Features

  • Real-time urban systems modeling
  • Physics-based simulations
  • AI-powered conversational queries
  • Integration with A.U.R.A. agent system
Unreal Engine Cesium Digital Twin AI Assistant 3D Tiles
Concept
Future Work

ATLAS: Local Voice Assistant for Project Management

"ATLAS," an on-device, local-first, multilingual (Telugu/English) voice assistant for construction knowledge management. Proposing an architecture using advanced STT/TTS models (e.g., Whisper, ViTs) to ensure privacy and offline functionality.

Proposed Capabilities

  • Hands-free project monitoring
  • Multilingual support (Telugu/English)
  • Offline, privacy-preserving operation
  • Real-time data entry and status updates
Whisper ViTs Local LLMs Voice Interface Multilingual

Interested in Collaboration?

I'm always open to discussing research collaborations, consulting opportunities, or speaking engagements related to construction AI, knowledge graphs, and intelligent document systems.

Get in Touch