DataRobot Downloads $54M Led by NEA to Automate Data Science Tasks
Jeremy Achin is embarrassed to admit it, but his decision to quit his job and start a company was inspired in part by seeing the movie “The Social Network,” the Hollywood take on Facebook’s origin story.
It was several years ago, and the then-31-year-old was a director of research and modeling at Travelers Insurance in Hartford, CT. But he was getting sick of working inside a big company with “lots of bureaucracy,” he says.
“I had an early mid-life crisis—what the heck am I doing with my life?” he recalls. “It had never occurred to me to start a company.”
One of the concerns was he had a mortgage and didn’t know how the startup would get funded. But he took a chance and left his job in 2012 to start a machine-learning software company, DataRobot, with Tom de Godoy, who also quit his data science job at Travelers.
They saw an opportunity in creating software to automate much of the work performed by data scientists. The idea is to make it easier for business users—even those without advanced technical skills—to quickly build and run predictive data models that might help boost decision-making, as well as their company’s bottom line.
“I just knew that the demand for data science and machine learning was going to really take off,” says Achin (pictured above), DataRobot’s CEO. “I just tried to picture what was going to satisfy that demand. Was it going to be hundreds of thousands of people like me, writing thousands of lines of code every day to solve these problems? I didn’t think that was likely.”
Funding the startup hasn’t been a problem. DataRobot has raised $57 million from investors over the past few years, and today it’s nearly doubling that with an additional $54 million equity investment led by earlier backer New Enterprise Associates. DataRobot is in talks with investors to expand the latest funding round by another $12 million to $15 million, Achin says.
That makes Boston-based DataRobot one of the biggest venture-backed technology bets in New England. And there is certainly a big business opportunity if DataRobot gets it right, says Rob May. He’s the CEO and co-founder of Talla, a Cambridge, MA-based startup developing virtual business assistants based on machine-learning technologies.
DataRobot is providing “a great service for this stage of the market, where there is a shortage of data science talent, but an increasing need,” May says in an e-mail. “I wish I was an investor.”
DataRobot’s software incorporates hundreds of machine-learning algorithms from open-source programming environments, such as Python, Apache Spark, R, and TensorFlow, Achin says.
Users input their data sets and tell the software what they want to glean from the data. The system “figures out the rest,” Achin says.
A banking customer might use the product to try and figure out who is likely to pay back a loan, while a healthcare client might try to figure out who is at risk for diabetes or heart disease, for example. DataRobot’s other customers include insurance firms, financial technology companies, manufacturers, and retailers, Achin says. (Side note: One of its early clients was the New York Mets, which used the software to help it pick players in the MLB draft, Achin says.)
Behind the scenes, DataRobot’s software “automatically runs a competition within itself by testing out hundreds or even thousands of solutions to the problem,” Achin says. It will then deliver the analytical models expected to provide the most accurate predictions. Achin says that aspect of the product was inspired by Kaggle, the Google-owned company that hosts competitions to crowdsource data science and machine learning tools. Achin and de Godoy, DataRobot’s chief technology officer, have both competed on Kaggle.com.
DataRobot has done a lot of work designing the front end of its product so that it’s simple enough for business analysts, executives, and others who aren’t trained in machine learning and statistics to use the software. The startup has also tried to make the details of the system’s operations “transparent” for customers, Achin says. “We don’t want to be a black box,” he adds.
Achin says the system can’t do everything a data scientist can, but it can solve many data modeling problems. “Our goal is to automate as much as possible,” Achin says.
Talla’s May says we’re a “long way from automating all the work of a data scientist, but, automating away the bottom 20 or 30 percent is probably very feasible.”
“If you have complicated data science needs, maybe this isn’t the right solution,” he continues, “but, if your needs are pretty straightforward, DataRobot can save you a lot of time and money compared to hiring a bigger data science team. I’m not that close to the company but, from what I know, it seems like a great idea.”
This is still a young and unpredictable sector, and May says one of the challenges for DataRobot (like most companies) will be navigating unanticipated obstacles.
DataRobot also has its work cut out differentiating itself from the myriad companies working on data analytics and machine learning platforms. Related companies include … Next Page »