Avi Patel is the 22-year-old founder of Kled AI, a San Francisco startup that buys and resells personal data from everyday people to train AI.
Patel dropped out of his studies after just two weeks in 2021 to become a founder. While running a music licensing startup, Patel was inundated with requests for AI training data — and decided to build a platform to provide it.
Today, Kled AI pays people to post videos of themselves taking out the trash or sending in photos of potholes on a road. This data is then sold on to robotics and self-driving car companies so they can improve their own AI.
While Kled AI is available worldwide, it’s particularly popular in countries like the Philippines, where an extra $20 to $40 a month just for photo uploads makes a bigger difference, Patel says.
The startup is among a new crop of AI training startups meeting the booming need for data to power the latest and greatest AI systems.
The following text is based on a conversation with Patel, which has been edited for length and clarity:
I dropped out of college after two weeks
I studied computer science at the University of Illinois, but I just didn’t like college. So I quit and immediately started a company.
Read too
My first business was in music licensing. We built a marketplace where artists could upload their music and companies could immediately license it. We had hundreds of millions of songs, but business wasn’t really going well. Dealing with record labels was simply too expensive.
At the same time, I kept receiving requests for data. Companies developing AI models asked if we had training data. That’s when I realized this was a much bigger opportunity.
Most of the data was collected without paying people
At first we tried a simple model – we bought content from production companies and resold it. That was boring. It was just a sales transaction.
So I thought: why not collect data directly from people instead? That’s exactly what Kled AI does. We developed an app where anyone can upload photos, videos or other data and get paid for it.
Most companies today collect your data without paying you for it. We’re trying to reverse that. People upload the data voluntarily and we pay them for it.
We ask people for very specific information
The key difference is that we don’t just collect arbitrary data. We assign tasks.
We might ask someone to make a video of themselves taking out the trash. This is useful for robotics.
We ask someone to take a photo of a food delivery at their doorstep. This helps with delivery and mapping systems.
Read too
Or we ask people to take two photos of the same scene – one with an object, one without. This is useful for image editing programs.
Because we control what users upload, the data is of much higher quality. Additionally, we only allow photos from people using newer iPhones.
The app has had great success in developing countries
We started about two months ago. The app reached number 1 on the App Store in Malaysia and became a hit in countries such as the Philippines and Indonesia. We now have over 200,000 users and receive around five million uploads every day.
Many of our users come from lower income countries. For example, in Malaysia, our average user earns $20 to $40 a month just by uploading photos occasionally. Our top users earn around $2,000.
We pay the same rates worldwide. I want everyone to have a fair chance.
Some users make this a full-time job
One of our top earners is a truck driver in the USA. He has multiple cameras on his truck and uploads footage throughout the day – plus photos of potholes, construction sites and other things he sees. This data is then resold to self-driving car companies and others.
He earns about $7,400 a month. He basically doubled his income just by uploading data. That’s exactly what I want to achieve – a way for regular people to benefit from the AI economy.
AI is only part of the market
Currently, most of our customers are AI and robotics companies. But I actually believe that AI is only a small part of the possibilities.
There is a much larger market among hedge funds, consumer brands and governments that are already spending trillions on data. Much of this data has been collected or extracted in the past without consent.
Companies would much rather pay for high-quality data collected with consent.
We just raised $6.5 million to expand this
We recently raised $6.5 million at a $150 million valuation. We are backed by Waymo founder Sebastian Thrun, French billionaire Bernard Arnault’s venture capital fund and others.
The funding will go towards building fraud detection systems – basically ensuring that the data uploaded is actually relevant to the task at hand. We already work with several leading AI laboratories.
Read too
I believe that at some point everyone will sell their data
Over time, I think this will become much more passive.
For example, you could link your Spotify or Netflix account and get paid based on usage. Hedge funds would pay a lot for such data. People could fund their subscriptions simply by sharing usage data.
At the moment we are a team of eleven people. But I think it can get much bigger.
For me, the goal is simple: instead of letting companies use your data for free, you should be able to make money from it.
