With the advent of social media, being a jerk on the internet has become easier than ever. Online moderators and profanity filters can only do so much to protect us from the inevitable vulgarity spewed all over these sites.
This is where Kite AI comes in. The startup, run by three college students, provides an API that uses machine learning to detect abuse and harassment. It helps to stop cyberbullying and other forms of online abuse to keep the internet that happy place it should be.
After seeing them on Product Hunt, I reached out and got to interview co-founder and CTO of Kite AI, Justin Potts.
How did the idea for Kite AI come about and how did you get involved?
The idea for Kite AI stemmed from an app idea my co-founder Alex Meza had, which was to create an app for people to rate fashion items (clothes, outfits, accessories…). He realized abuse and harassment would be a large problem on the platform, so started thinking of ways to prevent it. While Alex was thinking about his new idea, Trevor Nguyen, Alex, and I, bought tickets to HackGSU, in Atlanta, Georgia. We brainstormed what we might like to build, and Alex suggested his idea. We liked it, and were set on building an app for consumers to improve their browsing experience on Twitter. After building a prototype in 36 hours, we won the #HackHarassment Award. A few weeks went by afterwards, and we slowly started picking back up the development. Alex and I started working closely to develop a business model, and we pivoted from a consumer app to improve Twitter’s UX, to an API that anyone could add to their app. Today, Alex Meza is our CEO, responsible for the business model, sales, financials, and marketing. I’m the CTO, overseeing and contributing to the development of our API and some new projects we’re working on, and Trevor Nguyen is our VP of Engineering, responsible for the core API.
You just launched this August 2017! How long did it take the team to build Kite AI? And how has the launch been?
We started working on Kite AI on March 31st, and we launched on August 15th. You hear a lot about companies launching products very quickly (Twitter took two weeks to get a prototype out), but unfortunately the nature of machine learning (and especially an API others will be using) is that it takes a long time to optimize the model, and build the tools you’ll need to allow others to use the technology. There’s a lot that had to happen behind the scenes, and tons of iterations on our product, whether it was changing the landing page design and brand aesthetic, working with different platforms to get server costs lower while increasing the power/speed, or hand-training data for the model.
The launch has been great! We launched on Product Hunt and we were featured on the front page! We broke triple digits and now have over 130 upvotes I believe. We posted on a few other platforms as well, and we’ve seen a lot of users signing up for the platform, and we’ve received a lot of great feedback from the community on our product.
What makes Kite AI better than other similar services?
Kite AI’s most unique attribute is that we’re the only publicly available API that allows you to scan text for harassment or abuse. Google has a project called Perspective API that is similar to ours that detects the “toxicity” of phrases, but is mainly being used by media companies like the NYT or the Wall St. Journal for comment moderation. Additionally, while Google’s model is great at recognizing whether or not a comment may contribute positively to a conversation, it’s not great at actually detecting abuse. For example, saying something like “This is f*@king awesome” is marked as something like 98% toxic, whereas Kite AI is able to recognize while this may not be appropriate in all situations, it’s not abusive, and shouldn’t be classified as harassment. Lastly, we’re almost ready to release topic modeling. Not only will Kite AI be able to tell you whether it’s abusive or not, we’ll be able to tell you what kind, and what it might contain. This includes subjects like racism, profanity, sexism, religious discrimination, LGBTQ discrimination, political discrimination, and potential radicalization.
You and the other two members of the Kite AI team are in college. Has being in college aided you or distracted you from creating this service?
College can be a really great environment for startups and entrepreneurs, and we’ve seen a lot of talent come out of this age group. You can point to examples like Bill Gates, or Zuckerberg, but many others as well, like Sam Altman (President of Y Combinator), Travis Kalanick (former CEO of Uber), and others. However, college can also be very distracting. I’m taking 18 hours this semester, which for those of you who don’t know means that I go to class 18 hours every week, and have hours of homework for each class as well. Additionally, I’m heavily involved on campus as an executive for a fraternity, and have a software developer internship through an entrepreneurship organization. This means I’m learning how to juggle responsibilities between extra-curriculars, school, and work. College can also be really distracting through the obvious areas too. Parties, social events, football games, all take time away from Kite AI, and a large part of growing as an entrepreneur and a mature individual is learning how to say “no,” and overcoming “FOMO,” which is fear of missing out. I promise you, if you’re passionate enough about your idea and are willing to hustle to make it work, the benefit of hitting a home run will far outweigh the FOMO.
That being said, I wouldn’t say that college has distracted me too much in a negative way. I’ve learned where I need to focus my time and how to manage my priorities effectively, which are all really valuable skills I wouldn’t have learned without these distractions. However, if you’re not passionate enough or strong-willed enough to be willing to miss out on things you used to do, college is probably not a great environment for you to be starting a startup.
What has been the most difficult part of starting and running Kite AI?
Besides all of us being in school, the two most difficult parts have been learning the tech required to make what we’re doing possible, and training all the data to build our model. Machine learning and natural language processing is not only a fairly new trend in the tech industry, but also completely new to me and Alex. Luckily, Trevor is super smart, and has a huge interest in math so these concepts come more easily to him. To make these kinds of advanced models accurate and effective, they require huge amounts of data. We started by collecting the datasets ourselves, and hand-training these items, which was very time consuming. Now, we’re exploring alternative ways to collect and classify this data to make it much more efficient.
Lastly, do you have any words of wisdom for people interested in starting a tech startup?
Just get started. No excuses. All you need is a laptop and an idea. Of course, this comes with researching your idea, seeing who else is doing something similar, talking to your customers… Additionally, something I highly advise is to find a role model, whether it’s your mentor, or someone who is in a position you’d like to be in some day. If it’s the former, set up meetings with them, take them out to coffee, learn as much as you can and ask a bunch of questions. If it’s the latter, find all the interviews, talks, and blog posts you can. Study these, and think about traits and lessons you can apply to your life and your ventures.
You can follow Kite AI on their journey by:
< Checking out their website
< Reading their more thorough introduction on Medium
< Following them on Twitter