MemeFinder — AI-Powered Meme Search Engine

🔍

Text-Based Search Fails

Existing platforms rely on keyword tags and metadata. But memes are visual, contextual, and culturally layered — traditional search simply cannot understand them. You end up scrolling through hundreds of irrelevant results.

📱

Gallery Chaos

Your phone gallery has thousands of screenshots, memes, and photos all mixed together. Finding that one specific meme you saved three months ago means endlessly scrolling — there's no way to search by content.

⚡

Context Switching Kills Flow

In the middle of a chat, you need to leave the app, open a browser, search for a meme, download it, switch back, and attach it. By the time you find it, the moment has passed. The conversation has moved on.

MemeFinder is the answer.

We built an AI-powered mobile application that lets you search memes the way you naturally think about them — by describing the scene, the joke, or the feeling in plain language. Our semantic search engine, powered by multimodal AI embeddings, understands the visual and contextual meaning of 70,000+ memes and delivers the exact match in milliseconds. No tags. No keywords. Just natural language.

Core Technology

Built different.
Built to matter.

Four foundational technologies. Each one designed
to solve a real problem, not create a new one.

Global Semantic Search

The Ultimate
Meme Vault

Search through a curated, ever-expanding database of over 70,000 images and videos. Just describe the scene, the joke, or the vibe in plain English, and our AI retrieves the exact match in milliseconds.

FAISS Gemini Embeddings 70K+ Media

On-Device Local Search

Search Your Gallery.
Completely Secure.

Stop scrolling endlessly to find that one screenshot from three years ago. MEME-Finder indexes your local device gallery entirely offline, letting you context-search your own personal collection with zero privacy compromises.

On-Device Index Zero Cloud Upload 100% Offline

MEME-Finder Keyboard

Never Switch
Apps Again

Bring the power of AI search directly into WhatsApp, iMessage, or Instagram. With the native MEME-Finder Keyboard, you can search, preview, and drop videos, GIFs, and custom stickers right into your active chat.

Custom Keyboard Rich Media All Chat Apps

Serverless P2P Sharing

Off-the-Grid
P2P Sharing

Want to drop a massive meme folder to a friend in the same room? Bypass the cloud completely. Use our serverless peer-to-peer transfer to beam high-quality videos and images directly between devices at lightning speed.

Mesh Network No Cloud Lightning Fast

📲

Install MemeFinder

Download the app from App Store or Google Play. Enable the MEME-Finder Keyboard in your device settings for inline search.

💬

Describe What You Want

Type a natural language query like "cat standing like human" or "confused math lady". Our AI understands context, not just keywords.

🎯

Get Instant Results

FAISS-powered vector similarity search retrieves the most semantically relevant memes from 70K+ media files in under 200ms.

🚀

Share Instantly

Send directly from the keyboard into WhatsApp, iMessage, Instagram, or any chat app. No downloading, no switching apps.

Client Layer

Swift (iOS)

Kotlin (Android)

Custom Keyboard Extension

P2P

▼

Backend Layer

Java / Spring

Docker

Microservice Architecture

MinIO

PostgreSQL

▼

AI / ML Pipeline

Gemini Multimodal Embeddings

CLIP

FAISS Vector Similarity Search

Automatic Captioning (Gemini Flash)

Batch Embedding Pipeline

▼

Data Layer

FAISS Index (70K+ vectors)

Google Cloud Storage

On-Device CoreML Index

Semantic Search Pipeline

User queries and meme images are both converted into 768-dimensional embedding vectors using Google's Gemini multimodal model. FAISS performs approximate nearest-neighbor search across 70K+ vectors, returning the top-K most semantically similar results in under 200ms.

On-Device Privacy Engine

For local gallery search, we run a lightweight embedding model directly on-device using CoreML (iOS) and ONNX Runtime (Android). Your personal photos never leave your device — the entire indexing and search pipeline runs offline with zero cloud dependency.

Custom Keyboard Architecture

The MEME-Finder Keyboard is implemented as a native iOS Keyboard Extension and Android Input Method. It communicates with the main app via shared containers, enabling instant search, preview, and paste into any active chat without leaving the conversation.

P2P Transfer Protocol

We use Apple's MultipeerConnectivity framework and Android's Nearby Connections API to create a serverless mesh network between devices. Users can beam entire meme collections to friends at LAN speeds — no internet, no cloud, no file size limits.

Data Curation Pipeline

Our automated pipeline deduplicates and indexes memes from curated community sources. Each item is auto-captioned with Gemini Flash, embedded, and added to the FAISS index via a scheduled batch pipeline.

Scalable Microservice Infrastructure

The backend follows a microservice architecture built with Java/Spring, containerized with Docker. PostgreSQL handles relational data, MinIO provides S3-compatible object storage for media assets, and the entire system is designed for horizontal scalability.

Meet Our Team

We are passionate computer engineering students from METU, working together to revolutionize how people discover and share memes.

Hello! We are final year Computer Engineering students at Middle East Technical University (METU). MemeFinder is our capstone project that combines our passion for AI, web development, and internet culture.

Ahmet Kılavuz

Full-Stack Developer

kilavuz.ahmet@metu.edu.tr

Mithat Timurcan

Mobile Developer

mithat.timurcan@metu.edu.tr

Alperen Ovak

AI/ML Engineer

alperen.ovak@metu.edu.tr

B. Eray Pasinlioğlu

Mobile Developer

eray.pasinlioglu@metu.edu.tr

Adilhan Çetin

Backend Engineer

adilhan.cetin@metu.edu.tr

Prof. Dr. İsmail Sengör Altıngövde

Advisor

altingovde@ceng.metu.edu.tr

Search Memes
Like You Talk Feel Remember Describe Chat
About Them