Jorge Torres is the co-founder and CEO of MindsDB, a platform that helps anyone harness the power of machine learning to ask predictive questions about their data and get accurate answers from it. MindsDB also graduated from YCombinator’s recent Winter 2020 batch and was recently recognized by Forbes as one of America’s Most Promising AI Companies.
What initially attracted you to machine learning?
It is an interesting story. In 2008 I was living and working in Berkeley at a startup called Couchsurfing and saw this class (cs188- Introduction to AI). Although I was not affiliated with the university at the time, I asked Prof. John DeNero if I could sit in on a class and he would. This professor was brilliant and he really made everyone fall in love with the subject. It was the best thing that ever happened to me. I was amazed that computers could learn to solve a problem, I realized that this was fast and decided to make it my career.
There are several generational defining events in technology that only happen a few times in a person’s lifetime. I was lucky enough to witness the birth of the Internet, but too young to be anything more than a passive observer. I think Machine Learning is the next generation event, and I wanted to be a part of it in some meaningful way to advance technology and how we use it.
MindsDB started at UC Berkeley in 2018, can you share some insights into these early days?
UC Berkeley is one of the largest research institutions in the world and has a history of creating and supporting open source software, and we thought there was no better place to start with MindsDB. Our values aligned, they offered us our first check through the UC Berkeley Skydeck accelerator, and the rest they say is history.
The early days weren’t like many startups in the Bay Area; The only difference is that instead of working in a dusty Palo Alto garage, we were in the relative comfort of a Skydeck Penthouse co-working space (rent free).
I believe that there is enormous power in data. The more a company has, the more they can promote their business. But only if they can get meaningful insights from it.
In the fall of 2017, my best friend Adam Kerrigan (COO) and I came to the conclusion that too many businesses were facing limitations when it came to extracting meaningful information from their data. They realized that one of the biggest limitations is how many of these businesses are severely underutilizing the power of artificial intelligence. We believed that machine learning could make data and the intelligence it could provide accessible to everyone. That’s why we designed a platform that would allow anyone to harness the power of machine learning to ask predictive questions about their data and get accurate answers from it.
We call this platform MindsDB, and we’re focused on making it incredibly easy for developers to rapidly build the next wave of AI-powered applications that will change the way we live and work, and for businesses to mine their data.
Why did MindsDB focus on solving the data centric problem as opposed to machine learning centric?
If you look at the vast majority of AI research, a large percentage comes from academic institutions. ML has historically been model-centric because this is where research institutions can add perceived value; more research improves models or creates new ones, thereby producing better results. On the other hand, being data-centric, adding better quality/more relevant data to an existing approach is not easily publishable (a key KPI for researchers).
However, the vast majority of applied machine learning problems today benefit more from improved data than from improved models. This also aligns well with our mission to democratize machine learning, the vast majority of people outside of the Ml space don’t know much about ML, but they sure do know a lot about their data.
We saw that there are two types of companies, on the one hand, companies that are in the database, and on the other hand, they have not yet identified the databases, we understood that if the company is in the database group, then their data; the maturity already put them on the right track to really apply machine learning, while companies that hadn’t yet discovered databases still had a long way to go, so we focused on providing value to those who could actually pull it off.
How does MindsDB approach modeling and deployment in plain SQL?
We create model representations as tables that can be queried, so we effectively remove the concept of “location” from the picture. When you type CREATE VIEW in the database, that view is live when the command is processed, same as when you do CREATE MODEL in mindsdb.
People love MindsDB because of the simplification you bring to the ML-Ops lifecycle Why is simplifying machine learning deployments so important?
People love it because it eliminates unnecessary ETL pipelines, so fewer things need to be maintained. Our goal is to enable users to extract the value of machine learning without having to think about maintaining the ML infrastructure if they already maintain the data infrastructure.
What are some of the benefits and risks of being an open source startup versus a traditional startup?
An open source project can start with just an idea and people will help you build it along the way, with a close source approach you have to start with the same assumptions, but you better be right because no one will help you improve your product (at least not to the same extent, what open source), think of open source as a user-friendly approach to shared products.
MindsDB recently raised $16.5 million in Series A investment from Benchmark, why is Benchmark the perfect investor fit and how does their vision align with yours?
Benchmark has an impeccable track record in our industry, Chetan has helped companies like mongodb, elastic, airbyte become global leaders in their industry. We believe there is no better fit for MindsDB than Chetan and Benchmark Capital.
Thanks for the great interview, readers who want to know more should visit MindsDB.