Overview
Build an AI-powered restaurant recommendation service inspired by Zomato. The system suggests restaurants based on user preferences by combining structured restaurant data with a Large Language Model (LLM) to produce personalized, human-like recommendations.
AI Tools
Stitch by Google
Cursor
Objective
Design and implement an application that:
-
Accepts user preferences (location, budget, cuisine, ratings, and more)
-
Uses a real-world restaurant dataset
-
Leverages an LLM to generate personalized, human-like recommendations
-
Displays clear, useful results to the user
Brainstorming Product Architecture
As a designer, I approached the product by first defining the user inputs, budget, location, cuisine preferences, and dining intent. The architecture was designed to balance AI recommendations with relevant data retrieval.
The flow begins with user preferences collected through the frontend. On the backend, a dataset of 50K+ restaurants is filtered to narrow down the most relevant options. These filtered results are then sent to an LLM (Groq/GPT/Claude) through an API, where AI analyzes user intent and generates personalized recommendations. The final output presents the top 5–10 restaurants that best match the user's preferences.
This architecture helped me understand how UX, data filtering, AI reasoning, and system design work together to create a meaningful recommendation experience beyond traditional UI design.
Design Process
To bring the idea to life, I followed an iterative AI-assisted product development workflow that combined UX thinking with implementation. This process helped me move beyond traditional design deliverables and understand how product thinking, UX design, AI, and engineering come together to build real-world applications.
Building with Cursor
Used Cursor as an AI coding partner to translate product ideas into a working application. From generating code and implementing features to debugging issues, testing workflows, and refining functionality, Cursor accelerated the journey from concept to deployment while helping bridge the gap between design and development.
Final UI — Refined Experience
The DineAI interface was designed using Stitch, enabling rapid exploration of layouts and interactions. The experience focuses on helping users discover restaurants based on location, budget, cuisine, ratings, and personal preferences. The final design balances structured filtering with AI-powered recommendations, making restaurant discovery feel more personalized, transparent, and decision-friendly.
→ View app
My learnings
-
Designing with AI requires systems thinking, not just interface design.
-
Product architecture and UX are deeply connected in AI-powered experiences.
-
Rapid prototyping with AI tools enables faster experimentation, testing, and iteration.