Inspiration
🚧Construction projects 🚧 are notoriously complex, with a constant flow of documents, blueprints, emails, and photos from the field. For project managers and teams, keeping track of every update and finding critical information quickly is a daily struggle. Key information gets siloed in different platforms, and fieldwork often means communication gaps. We were inspired to create a centralized, intelligent hub that could cut through the noise, automate administrative burdens, and provide a single source of truth for construction projects.
What it does
Our platform, the concrete.work, introduces a Kanban board that intelligently populates itself by analyzing project files from Google Drive and incoming emails.
For example, when a site supervisor uploads a photo of a potential safety hazard to a designated Drive folder, our system automatically:
Analyzes the image using Gemini's vision capabilities to understand its content (e.g., "Image shows an unsecured ladder and water spillage on the floor").
Creates a new "Safety Concern" card on the Kanban board.
Populates the card with the generated image description, the file link, and other relevant metadata. Furthermore, we've integrated a powerful chatbot, powered by Letta. Any team member can ask questions in natural language like, "What are the latest change orders for the HVAC system?" or "Show me all recent photos related to plumbing," and receive instant, accurate answers drawn directly from the project's documents. This transforms document management from a manual chore into an on-demand, conversational experience.
How we built it
Our solution is a modern, full-stack application leveraging a suite of cutting-edge AI and web technologies:
Frontend Prototyping: We used v0 and Vercel to rapidly generate our user interface components. This allowed us to quickly prototype a clean, intuitive Kanban board and chatbot interface, ensuring a user-friendly experience from the start.
Backend and Hosting: The core of our application is a FastAPI backend running on Google Cloud Platform (GCP). We chose GCP for its scalability and robust ecosystem. The backend features secure connectors to the Google Drive and Gmail APIs, allowing our agents to access and process project data.
Autonomous Agent Technology: Letta is the engine of our automation. We built a backend that polls Google Drive and Gmail for new files and emails. This agent is responsible for the initial data ingestion and processing, handing off tasks to other services.
Conversational AI and RAG: Letta powers our intelligent chatbot. All textual data from Drive and Gmail is chunked, converted to vector embeddings, and stored in a vector database. Letta uses a Retrieval-Augmented Generation (RAG) architecture to search this database and provide contextually relevant answers to user queries.
Image Analysis: To bring visual data into the workflow, we utilized Gemini's vision capabilities, performing OCR and image recognition on non-text files.
Challenges we ran into
Our biggest technical hurdle was integrating both the frontend and backend within a reasonable amount of time. As a team, we were new to iterating and deploying these multifaceted aspects as quickly as possible. This involved understanding containerization with Docker, managing environment variables for our API keys securely, and setting up the correct IAM permissions for our service accounts.
Furthermore, the project required us to learn and integrate several new and advanced services within a very short timeframe. Getting, Letta's RAG system, and Gemini's vision API to communicate seamlessly with our core Next.js application was a significant but rewarding challenge.
Accomplishments that we're proud of
We are incredibly proud of successfully building a functional, end-to-end prototype that demonstrates a powerful new paradigm for construction management.
What we learned
This project was a tremendous learning experience. On the technical side, we gained hands-on experience with deploying serverless applications, working with autonomous agent technology, and implementing a complex RAG pipeline. We learned the importance of robust API security and the nuances of cloud service configuration.
More broadly, we learned how a combination of different AI technologies can be orchestrated to solve a real-world business problem. We saw firsthand how moving beyond simple chatbots to a more agentic, proactive system can unlock significant value and efficiency.
What's next for 🏗️ concrete.work 🏗️
We believe our platform has the potential to become an indispensable tool for the construction industry. Our roadmap for the future:
- Real-time Notifications: Implementing push notifications to alert project managers to urgent updates, such as newly identified safety concerns or critical change orders.
- Expanded Data Source Integration: Adding support for other industry-standard data sources, such as Procore, Autodesk Construction Cloud, and various scheduling software.
- Predictive Analytics: Leveraging the collected project data to train machine learning models that can predict potential project delays, cost overruns, and safety risks before they happen.
Log in or sign up for Devpost to join the conversation.