<aside> 📌
proofread by ai; should not have mistakes
</aside>
Amazing folks at Soma Capital sponsored by AWS and Anthropic had organized a one day NYC AI Hackathon this weekend at their Soho office (truly beautiful).
I participated in a team of three with Naomi Rufian (Cornell) and Aneesh Maganti (NYU), where we built Savor - an intelligent restaurant discovery platform that uses AI to find the perfect dining spot for any occasion. What started as a simple idea (as one of us had date and were searching for restaurant) turned into a fully functional web application.
Finding the right restaurant for a group can be surprisingly complex. You need to consider dietary restrictions, budget constraints, atmosphere preferences, proximity, group size and whole bunch of unique problems - all while ensuring everyone has a great experience. Traditional restaurant apps often require extensive filtering and manual comparison, which can be time-consuming and overwhelming, yet still not give you best results or experience.
Having options is common in NYC, but too many options is decision making nightmare
Savor takes a different approach. Instead of overwhelming users with endless options, it collects their preferences through an interactive agentic chat and uses Anthropic's Claude AI and OpenAI’s GPTs to intelligently match them with the perfect restaurants. The AI doesn't just recommends restaurants - it explains why each restaurant is a great match for their specific needs. Why scroll through 100s of good restaurants when you can get the one that fits your requirement directly.
<aside> 📌
Sample Input: I’m going out on a dinner date with budget not exceeding 100 for two, none of us will drink (bummer?) and would prefer to try out some mexican food in or near midtown.
</aside>
Savor recognizes the occasion and finds out ideal date location so that it’s not too noisy or crowded, while also making sure you don’t travel long, and probably less/no drunk people around.
Smart Preference Collection: The form captures everything from dietary restrictions to atmosphere preferences, ensuring comprehensive matching criteria.
AI-Powered Matching: Claude and GPT powered n8n automated workflows analyzes each restaurant against user preferences, providing match scores and detailed reasoning for why each recommendation fits.
Real-Time Results: Users receive 5-8 curated recommendations instantly, complete with reservation links and AI-generated explanations.
Mobile-First Design: Responsive design ensures the app works seamlessly across all devices.
The heart of Savor lies in its AI integration. When a user submits their preferences, the app structures this information into a detailed prompt for Claude, which then: