AutoTinder

Inspiration

I wanted to learn n8n cause the automation service is cool, and this seemed like a task that I personally needed to get done so it was a pretty good learning experience.

What it does

Simulates human activity and swipes based on my personal preferences at random intervals. Random amounts of swipes at random times throughout the day. Once a match is detected, it goes through a question bank to get the conversation going. When the conversation deviates from the decision tree the automation sends me an SMS message to take over. Also, when n8n detects a date and time in the chat, it will automatically add it to my Google Calendar.

How we built it

I swiped on Tinder for 4.5 hours labeling profiles with Yes/No/Unsure (unsure means there's multiple people in the image or we can't see a face, etc.) and used that as a logistic regression layer on top of OpenAI's CLIP Model. It was able to do all the swiping without any human assistance so then I hooked it up to n8n and then created the rest of the workflow.

Challenges we ran into

It needed way more data than I thought it would. The 4.5 hours of swiping and labeling turned out to only be like 1/3 of the data it needed to be accurate enough. Over time I added more data and it eventually fixed itself.

Accomplishments that we're proud of

It works

What we learned

n8n

What's next for AutoTinder

Set up a Model Context Protocol (MCP) server based on all my downloaded iOS text messaging history so that when the bot is messaging matches on Tinder it simulates my exact texting style.

Built With

• joblib
• numpy
• openai-clip
• pyautogui
• python
• pytorch
• scikit-learn