AI Product Decision Copilot

Inspiration

Choosing a smartphone today is overwhelming. Users are forced to compare endless specs, reviews, and marketing claims without knowing what truly matters for their needs. We wanted to build an AI assistant that thinks like a knowledgeable product expert—understands user intent, reasons over trade-offs, and gives a clear recommendation instead of a comparison overload.

What it does

AI Product Decision Copilot helps users discover the best smartphone based on their budget, usage type, and priorities. Users enter simple preferences, and the app returns:

The best-matched device with confidence and reasoning
Clear pros and cons
Two alternative recommendations
A concise summary to support decision-making
All results are personalized and easy to understand.

How we built it

The frontend is built using React, Tailwind CSS, and modern component-based architecture.
User inputs are sent as structured JSON to a Node.js + Express backend, which integrates with the Google Gemini 3 API.
We designed prompt templates that guide Gemini to reason over constraints and return strictly formatted JSON responses, making the output reliable and UI-friendly.

Challenges we ran into

Ensuring Gemini responses were consistently valid JSON
Selecting the correct Gemini model and API version
Designing prompts that balance reasoning depth with predictable output
Handling asynchronous AI responses while keeping the UI responsive

Accomplishments that we're proud of

Built a full end-to-end AI-powered product in a short timeframe
Achieved clean separation between UI, backend, and AI logic
Designed a user-friendly interface that feels simple yet powerful
Successfully leveraged Gemini 3 for real-world decision reasoning

What we learned

We learned how prompt design directly impacts AI reliability, how to integrate large language models into production-style applications, and how to structure AI outputs for frontend consumption. We also gained experience handling real-world AI edge cases and latency.