The process of forming a new software development team is often more art than science. We spend countless hours matching technical skills, but often neglect team chemistry and cultural fit. That’s what inspired the Dream Team Assembler—a tool designed to use AI to score compatibility between a project’s needs and a potential team’s attributes.
Building this system was a fantastic exercise in serverless architecture, primarily leveraging Google Cloud Run.
Why Cloud Run?
I needed two distinct services: a React frontend to host the user interface, and a robust backend API to handle the heavy lifting of AI processing and data persistence.
Frontend Containerization: Instead of relying solely on Firebase Hosting, I chose to deploy the entire React application in a container and serv…
The process of forming a new software development team is often more art than science. We spend countless hours matching technical skills, but often neglect team chemistry and cultural fit. That’s what inspired the Dream Team Assembler—a tool designed to use AI to score compatibility between a project’s needs and a potential team’s attributes.
Building this system was a fantastic exercise in serverless architecture, primarily leveraging Google Cloud Run.
Why Cloud Run?
I needed two distinct services: a React frontend to host the user interface, and a robust backend API to handle the heavy lifting of AI processing and data persistence.
Frontend Containerization: Instead of relying solely on Firebase Hosting, I chose to deploy the entire React application in a container and serve it via a dedicated Cloud Run Frontend Service. This demonstrated a fully containerized deployment pipeline, which is highly scalable and portable.
Stateless API for AI Orchestration: The core logic—taking a project description and asking the Gemini API to decompose it into roles, or calculating a complex “Chemistry Score”—is handled by a Cloud Run Backend API. Cloud Run’s rapid scale-to-zero capability is perfect here. The AI calls are bursty; the service only needs to be active for a few seconds to process the request, and then it can immediately scale back down, keeping costs minimal.
AI is the Secret Sauce
The real magic happens with the Gemini 2.5 Flash API. This model is integrated into the backend to execute the two most critical functions:
Project Decomposition: By feeding the model the project scope, it returns a structured list of necessary roles (e.g., “3 Frontend Engineers,” “1 Cloud Architect”).
Chemistry Scoring: This is more complex. I feed Gemini the project’s non-technical requirements (e.g., “fast-paced,” “high autonomy”) alongside candidate profiles and prompt it to return a normalized 1-100 score and a brief justification. It acts as an incredibly fast, highly nuanced compatibility expert.
Key Takeaways
Working with Cloud Run and Gemini provided three major takeaways:
Seamless Integration: Calling the Gemini API from a Cloud Run service is straightforward, turning complex analytical tasks into a simple API call within my business logic.
The Power of Statelessness: Cloud Run forces a clean, stateless design. This discipline is excellent for reliability, as every request is independent, and the service can scale horizontally without effort.
Maximum Platform Utilization: For a submission, using Cloud Run for both the frontend and the AI-powered backend is a powerful way to demonstrate mastery of the serverless ecosystem.
This project proved that advanced AI features can be integrated into powerful, scalable web applications without needing to manage any infrastructure. The future of high-impact software is truly serverless!