Project Overview & Award
🏆 1st Place Winner - $4,000 Scholarship
Awarded at the ACRES Competition, Thompson Rivers University (December 2024) for exceptional innovation and project execution.
As Team Lead, I spearheaded a team of three talented TRU Computing Science students (Deeparsh Singh, Khushal Arora, and myself) to develop the ACRES AI Construction Management Platform. This project aimed to revolutionize construction site operations by integrating cutting-edge AI for enhanced safety, efficiency, and project coordination.
We designed and built a comprehensive web application featuring automated uniform/PPE compliance checks using computer vision, an intelligent AI chatbot for real-time data access via secure SQL queries, and dynamic task generation and management leveraging AI agentic calls.
The Industry Challenge
The construction industry, while vital, often faces significant operational hurdles. Traditional site management practices can lead to inefficiencies and safety risks. Key challenges we identified include:
Safety Compliance Gaps
Manually ensuring consistent Personal Protective Equipment (PPE) usage across diverse sites is difficult, time-consuming, and prone to human error, increasing safety risks.
Inefficient Data Access
Tracking inventory and accessing critical site data often relies on manual methods or siloed systems, hindering real-time decision-making and resource allocation.
Complex Task Coordination
Managing project tasks, dependencies, and progress updates manually requires significant overhead and can lead to delays or missed steps, especially in dynamic environments.
Our AI-Driven Solution
To address these challenges, we developed the ACRES Management Platform – an integrated system leveraging AI to streamline workflows. The platform centers around three core functionalities designed to work seamlessly together:
Core Pillars
- Automated Uniform Compliance: Using AI vision to verify PPE adherence automatically.
- Intelligent Inventory Management: An AI chatbot providing secure, real-time SQL database access via natural language.
- Dynamic Task Breakdown: AI agents generating and organizing project tasks within a Kanban system based on context.
Feature: Automated PPE Compliance
Enhancing worker safety was a primary goal. We implemented an AI-powered system to automate the crucial process of checking for correct Personal Protective Equipment (PPE).
Site-Specific Configuration
Site managers use a dedicated dashboard to define the mandatory PPE items (e.g., hard hats, safety vests, steel-toe boots, gloves) required for each unique construction site. This allows for granular control based on site hazards.
Managers can easily add or modify requirements, including custom gear, ensuring the system stays up-to-date with safety protocols.
AI-Powered Verification
Workers simply capture an image of themselves upon arrival. Our integrated computer vision model analyzes the photo in real-time, detecting the presence of required PPE items.
The system provides immediate visual feedback, highlighting detected items (e.g., in green) and missing items (e.g., in red), enabling quick corrections before entering the work zone.
Feature: Intelligent Inventory & Data Chatbot
Accessing up-to-date information quickly is crucial for efficient operations. We built an AI chatbot, the "ACRES Database Assistant," to provide a natural language interface for querying the project's operational database securely.
Conversational Data Access
Users can ask questions like "How many employees do we have?", "What materials are needed for the Uptown site?", or "Show me the inventory level for concrete."
The chatbot understands the intent and retrieves the relevant information from the database.
Secure SQL Integration
The chatbot doesn't just understand language; it translates user requests into precise SQL queries executed against the database backend. This ensures data integrity and allows for complex data retrieval involving multiple tables (e.g., finding nearby sites with specific inventory).
Crucially, the raw database structure and data are never exposed via external APIs. The AI handles the interaction internally, enhancing security.
Data and Results
Feature: Dynamic Task Breakdown
Streamlining project management and task delegation was another key focus. We implemented a Kanban-based system enhanced with AI agentic capabilities to intelligently generate and organize project tasks.
Visual Task Management
The Kanban board provides a clear visual overview of tasks categorized by status (To Do, In Progress, Done). Each task card includes details like priority, due dates, descriptions, subtasks, and specific instructions, facilitating easy tracking and updates.
AI-Powered Task Generation
Managers can leverage an "AI Agent" to automatically draft a set of tasks for a specific job or phase. By providing context – such as the type of work (e.g., "water pipeline installation"), relevant site information, quality assurance documents (ITPs), and the current state of the board – the AI generates a structured task list.
AI-generated draft Kanban for "Water Pipeline Installation".
Intelligent Delegation & Planning
The AI analyzes the inputs and breaks down the work into logical, actionable tasks and subtasks (e.g., Plan Route, Acquire Permits, Procure Materials, Prepare Site, Schedule Team). This saves significant planning time and ensures comprehensive coverage of project requirements, minimizing the risk of overlooking critical steps.
Technical Deep Dive
Building this platform involved integrating several key technologies and overcoming interesting technical challenges:
AI & Machine Learning
- Computer Vision (PPE Detection): Utilized pre-trained object detection models, fine-tuned on construction site imagery, to accurately identify various PPE items in diverse conditions. Addressed challenges related to lighting variations and partial occlusions.
- Natural Language Processing (Chatbot): Leveraged Large Language Models (LLMs) for intent recognition and entity extraction from user queries. Developed robust prompt engineering techniques to reliably translate natural language into structured SQL commands.
- Agentic AI (Task Generation): Employed an agentic framework where the AI could reason based on multiple sources of context (manager input, ITP documents via RAG - Retrieval-Augmented Generation, current Kanban state) to generate relevant, structured task outputs in JSON format suitable for the Kanban board.
Backend & Database
- SQL Database: Designed a relational database schema to efficiently store and relate data for sites, employees, materials, inventory, PPE requirements, tasks, and quality documents.
- Secure Data Handling: Implemented backend logic to ensure the AI chatbot interacted with the SQL database securely, preventing SQL injection and avoiding direct exposure of database credentials or structure to the frontend or external services.
Frontend & Architecture
- React & TypeScript: Built the user interface using React and TypeScript for a type-safe, component-based, and interactive experience. (Likely leveraged Next.js, specify if confirmed).
- State Management: Employed appropriate state management solutions (e.g., Context API, Zustand) to handle application state, especially for the dynamic Kanban board and chatbot interactions.
- API Integration: Developed APIs to connect the frontend with backend services handling AI model inference, database queries, and task management logic.
Results & Recognition
Competition Success
The ACRES AI Construction Management Platform was awarded **First Place** in the ACRES Competition at TRU (December 2024), along with a **$4,000 scholarship**. This recognition validated our innovative approach, the technical sophistication of our solution, and its potential impact on the construction industry.
The judges highlighted the seamless integration of multiple AI technologies to solve real-world problems in safety, data access, and project management.
Key Learnings & Takeaways
Leading and developing this project was an invaluable experience, offering significant learning opportunities:
- Leading a Technical Team: Gained practical experience in project planning, task delegation, managing timelines, and fostering collaboration within a development team to deliver a complex product under competition deadlines.
- Integrating Diverse AI Systems: Learned the intricacies of combining different AI modalities (vision, NLP, agentic reasoning) into a cohesive application, including managing dependencies and data flow between them.
- Bridging AI and Databases Securely: Mastered techniques for enabling AI models (like LLMs) to interact with structured data (SQL) in a secure and effective manner, focusing on preventing vulnerabilities while maximizing functionality.
- Full-Stack Development Practice: Reinforced skills across the entire development stack, from designing intuitive frontend interfaces with React/TypeScript to building robust backend APIs and managing database interactions.
- Problem Decomposition & Solution Design: Honed the ability to break down complex real-world problems into smaller, manageable components and design targeted technical solutions for each, ensuring they integrate into a functional whole.