Campaign: Built as part of the Google Cloud Gen AI Academy APAC — "Meet the Builders" campaign.
Introduction
Hello community! My name is Ananya Raj, and I am a 2nd-year B.Tech Computer Science and Engineering student. Like most tech students, I love exploring new technologies and regularly use various free AI tools to make my coding, studying, and daily work easier.
However, as AI rapidly evolves from simple chat interfaces to autonomous, production-grade agents, I wanted to take a major step forward. I wanted to escape the "it works on my machine" loop, move my code out of isolated Jupyter notebooks, and build a real-world application that can be seen and reviewed by Google Cloud experts.
That is why I joined Cohort 3 of the Gen AI Academy APAC Edition! For my very first task, I am thrilled to introduce Marketera—an intelligent, safe, and conversational marketplace designed specifically for university campuses.
The Campus Problem: The Chaos of Chat Groups
On any college campus, students are constantly buying, selling, or sharing resources—whether it's second-hand engineering graphics textbooks, lab coats, drafters, mattresses, or cycles.
Currently, this campus trading relies on disorganized WhatsApp groups, Telegram channels, or local forum threads. This creates massive friction:
- Search Fatigue: Students have to scroll through hundreds of chaotic messages and spam to find a single relevant item.
- No Structured Filters: Finding something with a specific budget, condition, or pickup location (e.g., "near the library" or "outside Hostel 3") is almost impossible.
- Moderation Nightmare: Campus groups quickly get flooded with off-topic spam, commercial ads, or duplicate listings, making them hard to trust and maintain.
The Solution: Enter Marketera
Marketera is a smart campus marketplace assistant powered by Generative AI. Instead of wrestling with complicated filters, search terms, or endless chat messages, students can describe exactly what they need in plain, everyday language.
For example, a student can type:
"I need a second-hand engineering graphics textbook and a drafter in good condition under ₹700, preferably near the library."
Marketera acts as an intelligent coordinator: it understands the request, extracts the constraints, searches active campus listings, ranks the matching items, and provides a friendly explanation of the recommendations.
Furthermore, to keep our campus environment clean and professional, Marketera integrates an automated Campus-Moderator layer to screen out spam, unauthorized off-campus listings, and inappropriate content before they ever go live.
How Marketera Works (Step-by-Step AI Workflow)
Marketera handles student requests through a highly structured, agentic workflow:
- Natural Language Input: A student enters their query or list of items in conversational text.
- Intent & Constraint Extraction: Google Gemini processes the query to identify key entities: item category, required condition, maximum budget, and preferred campus pickup location.
- Database Querying: The backend searches active student listings stored in Google Cloud Firestore.
- Intelligent Ranking & Moderation: The system filters out closed or irrelevant items, ranks matches based on price and location proximity, and screens listings against our campus safety guidelines using the moderator layer.
- Conversational Explanation: Gemini generates a clear, personalized response outlining the best options and explaining why they fit (e.g., "This graphics textbook is available for ₹500 from a senior near Hostel 3, well within your budget").
- Direct Connection: The buyer can instantly connect with the student seller to coordinate the handoff safely on campus.
Our Google Cloud Developer Stack
To move Marketera from a local prototype to a robust cloud application, I leveraged Google Cloud's developer ecosystem:
- Google Gemini API (
gemini-2.0-flash): The primary AI engine used for natural language intent extraction, semantic item matching, and generating user-friendly responses. - Google Cloud Run: Hosts our containerized backend API. Since Cloud Run is fully serverless, it scales down to zero when students are in classes or asleep, making it exceptionally cost-efficient to run.
- Google Cloud Firestore: A scalable NoSQL database used to store active student listings, seller profiles, and campus pickup locations in real time.
- Google Cloud Secret Manager: Securely stores our Gemini API keys and credentials, preventing leaked-key billing surprises.
- Google Artifact Registry & Cloud Build: Powers our CI/CD pipeline, automatically building our Docker container images and deploying them to Cloud Run directly from GitHub.
- Cloud Logging: Logs active searches, moderation events, and performance metrics to help debug and optimize the application continuously.
Key Takeaways & What's Next
Building Marketera taught me that the true power of AI lies in how we deploy it under real-world constraints. By transitioning from standalone prompting tools to architecting serverless cloud microservices on Google Cloud, I’ve learned how to design secure, scalable, and highly practical applications.
My next milestone is to finalize the deployment, secure feedback from the Gen AI Academy mentors, and pitch Marketera in the upcoming cohort Ideathon!
Are you a builder in Cohort 3? What local challenges are you tackling with Gemini? Let’s connect in the comments below!
Top comments (0)