Cloud Engineer /
DevOps Engineer

Key components of the solution:

• Vertex AI Agent Engine: Hosts the Gemini-powered agent logic built using ADK.
• Image & Audio Generation: Uses Imagen 4 (Preview) for visual aids and Chirp3 TTS for local-language stories.
• Speech-based Assessment: Uses STT to evaluate student fluency.
• Vertex AI Search: Retrieves answers from NCERT textbook content ingested into Discovery Engine.
• Deployment: Frontend in Streamlit, Dockerized, deployed via Cloud Run. Content stored using Firebase + Cloud Storage.

Sahayak brings AI-powered innovation to underserved classrooms with a strong focus on contextual, accessible education.

The architecture was designed for scalability, performance, and security using various GCP services:

• Static Website Hosting (GCS): Website files (HTML, CSS, JS, images) are stored in a Google Cloud Storage bucket configured for public access and versioning.
• Content Delivery (Cloud CDN): Cloud CDN caches content globally, ensuring low latency and fast load times for users worldwide by serving content from edge locations close to them.
• Traffic Management & Security (HTTP(S) Load Balancer): An HTTP(S) Load Balancer acts as the entry point, managing incoming public web traffic, providing a custom domain name, and handling SSL/TLS termination for secure HTTPS connections via Google-managed certificates.
• Dynamic Functionality (Cloud Functions & Firestore): A dynamic visitor counter was implemented using a serverless approach. A Node.js Cloud Function is triggered on page load, interacting with Firestore (a NoSQL database) to atomically increment and retrieve the visitor count, which is then displayed on the frontend.
• Automation (Cloud Build & GitHub): A CI/CD pipeline was established using Cloud Build, triggered by pushes to a GitHub repository. This automates the build and deployment process, ensuring the latest code changes are seamlessly deployed to the GCS bucket.

This multi-service approach demonstrates proficiency in leveraging different GCP components to create a well-architected cloud solution.

The development of Craft a Story was approached with a strong emphasis on building a scalable, easily maintainable, and exceptionally user-friendly application. Several key architectural decisions were made to ensure these objectives were met:

• Flutter Frontend: The user interface was built using the Flutter framework to leverage its cross-platform capabilities, allowing the app to seamlessly operate on both Android and iOS devices. Flutter's rich UI toolkit enabled the creation of a visually appealing and intuitive user experience.
• Firebase Backend: To manage the application's data and business logic, a robust backend infrastructure was established using Firebase services.
  • Firebase Firestore was utilized for real-time data storage, providing a scalable and flexible NoSQL database solution.
  • Firebase Cloud Functions were implemented to create a serverless API, responsible for handling user requests, managing authentication, and orchestrating the AI story generation process.
  • Firebase Storage was used to efficiently manage and serve the application’s media assets, including images, audio, and video files.
• AI Integration: At the heart of Craft a Story lies its intelligent AI engine, which leverages powerful Generative AI models to craft unique and compelling stories.
  • Gemini 2.0 by Google, was used to ensure high-quality output. This allowed us to personalize plots and characters according to user specifications, providing each child with a truly unique and imaginative narrative.
  • High-quality voice generation was achieved by integrating Google Cloud Platform's Text-to-Speech (GCP TTS) service, offering a broad spectrum of realistic voices across multiple languages to enhance the storytelling experience.
• Image Generation: The integration of the Flux image generation model enabled the generation of compelling images to tell the story. The high quality allowed for stories with unique artistic styles.