While preparing for the Google Cloud Architect certification, I wanted to deepen my understanding by building something practical using the GCP services I was learning. That led me to start a news summarization application, this is something I've been developing and refining in my personal time to apply cloud and AI/ML concepts in action.
My goal was clear, while working on app to create a fast, minimal, and clean reading experience that offers summarized news instead of going through the large articles, giving readers the key updates of the articles without scrolling endlessly. It is a mobile-first, web-based app designed for easy to use. Readers can swipe through articles just like they would do while browsing reels or short videos. This makes web based app experience intuitive and similar to mobile natives app.
Architecture of application
At the same time, I chose Next.js to build the app, which made it easier to deploy on Cloud Run and helped to meet the SEO requirements effectively. As we know, strong SEO is essential for any web-based application, and Next.js with its built-in server side rendering and dynamic routing perfectly supports these needs.
Over time, the technical scope expanded, allowing me to explore and apply a wide range of GCP Services. Here’s a quick breakdown of what I have used so far:
- Firebase : for app authentication, Firestore for NoSQL, Hosting, and Functions for backend logic
- Cloud Run : to deploy the Next.js app.
- Cloud Functions & Cloud Scheduler : to fetch and summarize news at regular intervals
- Cloud Storage and Cloudinary : to manage article media.
- Secret Manager & IAM : to secure resources access.
- Artifact Registry : to store artifact after build.
- Cloud Build : to build the project
- Analytics and Logs : for monitoring
- Next.js (using TypeScript) : chosen for its server-side rendering capabilities and SEO friendliness.
- TailwindCSS : to create a smooth and responsive user experience
- Other : Swiper.js, etc
- Using a vector database to enable similarity-based retrieval
- Integrating with a retrieval-augmented generation (RAG) to give fast response to user-queries.
- Working on fine-tuning the T5-small model using the CNN/DailyMail dataset test data for summarize the articles. This will take lot of time and I am going to post learning from this fine-tuning as well.
- And many more...
You can explore the app here(Mobile-first) : https://www.inbits.co
(Feedback is always appreciated!)
This entire experience has been a way for me to stay hands-on, apply what I’ve learned, and keep up with real-world use cases in cloud, full-stack development, and machine learning.
Top comments (0)