Jason Ge, Founder of Snark AI, on Putting the “Intelligence” in Artificial Intelligence
Greetings beloved readers,
Welcome to another edition of PIE, where we interview the brightest entrepreneurs and VCs from Princeton and beyond. This issue’s featured founder is Jason Ge, a graduate student at Princeton from 2014–2018 and founder of Snark AI. The SaaS (software as a service) startup aims to build a software layer on top of all the computation resources for deep learning. They help their clients train cutting-edge deep learning models on terabytes of data, and with their technology, training time can be reduced from months to days.
1. How did the idea for Snark AI originate?
While my co-founder Davit and I were doing research at Princeton, we found out that the ecosystem for deep learning on cloud is very immature. It was very costly and difficult to train and deploy large scale deep learning on cloud. That inspired us to start a company to solve the problem.
2. Did you receive help along the way when starting the company and in its initial stages? If so, from whom?
Yes, Davit and I received a lot of help. In the initial stages it was mostly from our PhD advisors at Princeton, Prof. Mengdi Wang and Prof. Sebastian Seung. They were very supportive of us pursuing the startup idea as a side project.
3. How did you secure venture capital funding for Snark AI? Do you have advice for early stage companies looking for funding?
We got into the startup incubator Y Combinator which gave us $120K as a pre-seed investment. YC helped us grow the company over the summer this year and organized the YC Demo Day when they invited a lot of VCs to invest in YC companies. There were a lot of VCs and angel investors looking to invest in YC companies and deals were closing very fast after Demo Day. YC created the FOMO dynamics round Demo Day which was great for helping us raise a few million dollars venture capital.
4. What was the moment or series of events that made you realize that this company was worth betting your career?
The moment we were accepted into Y Combinator. It was a big deal for us.
5. What unforeseen challenges did you encounter in this whole process?
We started the business by letting crypto-mining farms rent out cheap GPUs to AI companies for deep learning. Crypto-mining GPUs were much cheaper than GPUs on public cloud because crypto-mining is not very profitable these days. But it turned out that companies were concerned about the data security problem and in general not ready to use crypto-mining GPUs. We had to move up the stack and build software that supports deep learning on all kinds of GPUs resources, including public cloud and crypto-mining GPUs.
6. What experiences prior to founding Snark AI inspired you to go into your field and contributed the most with developing your entrepreneurial skills?
Reading Paul Graham’s essays. Watching YC startup online courses. Working on open-ended problems. My co-founder Davit and I worked together on Alexa Prize Competition in 2017 to build a chatbot on Alexa which can have conversation with real person for 10 mins. We got the chance to know each other during the competition and we then decided to start the company together.
7. What is the most valuable piece of advice you would give to aspiring entrepreneurs?
Validate and iterate your idea with real clients. Avoid being judgmental on any idea in the early stage because the best startup ideas are usually counter-intuitive — they don’t seem to be good ideas in the beginning but will turn out to be great if executed well.
—
This interview is brought to you by Allen Liu ’22. Reach out to him here for questions or comments. If there’s anyone you’d like for us to interview just let us know and we’ll get right on it :)
Check out our other interviews here!
