SiftScamDetector

Inspiration

In the life we all live in , millions of people use platforms like Facebook Marketplace, Craigslist, and OfferUp to buy and sell items. But with that convenience comes a dark side: scams. After seeing friends and family fall victim to fake listings, we wanted to build something that could give users more confidence and safety when shopping online. That’s where Sift was born—a tool that uses AI to help everyday people quickly assess the trustworthiness of an online listing.

What it does

Sift is an AI-powered scam detection web app for peer-to-peer marketplaces. Users can input a link to a listing or upload an image, and Sift analyzes it across three dimensions: the text description, the price, and the images. It then generates a “scam score,” highlights red flags (like vague language or stock photos), and even suggests safer, verified alternatives. The goal is to make spotting scams as easy as clicking a button.

How we built it

We built Sift using the latest Next.js App Router, with React and TypeScript for a fast and scalable frontend. We used Tailwind CSS and the shadcn/ui component library for a clean, responsive design. For AI analysis, we integrated OpenAI’s GPT-4o for text and image inspection. Supabase powers our backend database, storing scan results, red flags, and alternatives. We also used Vercel Blob for image uploads and hosted the entire app on Vercel for seamless CI/CD.

Challenges we ran into

Some specifc challenges we ran into were the sturcutre of design itself because to design such a thing to return JSON from Gpt-4o was a bit hard and confusing especially for vision- based image inputs. Furthermore, was the trouble witht eh image upload flow because making the uploads smooth and consistent took a bit of effort than we excpected but we did learn how to use Vercel Blob. We had to prioritize the creation of the actual tool than other thing like security features such as user acess authentiation

Accomplishments that we're proud of

We made a working app in a couple of months with the use of SupaBase and React Native for the base of the web app

What we learned

We learnt how to use new tools such as Supabase and Vercel Blob

What's next for SiftScamDetector

Creation of an user authentication page to keep an profound security for the web app maybe we will sue E2E(end to end) encryption to keep it secure.

Built With

• openai
• react
• react-native
• supabase
• tailwindcss
• typescript
• vercel
• vercelblob