It’s a great time to be a software developer. And to be using artificial intelligence to solve real-world problems. Combine those trends, and you have a company called Beansprock.
The Boston startup is rolling out a new site today for software engineers looking for jobs. The idea is you sign up, tell the site what kinds of positions and companies you’re interested in and how actively you’re looking, and turn things over to Beansprock. The company then sends you e-mails with the best job matches it can find over the course of days and weeks.
Beansprock is led by two MIT graduates, Cameron Levy and Dustin Smith. Levy is the business guy (pictured above on left). Smith is the tech guy (on right). They incubated their company at the MIT Media Lab through a program called the E14 Fund; Beansprock was one of six startups in E14’s first class, which ran from September 2013 through April 2014.
The name Beansprock is meant to convey the idea of climbing a beanstalk—like a corporate ladder—and also sprockets, gear-like structures that signify a good fit, says Levy, an MIT Sloan School alum who serves as CEO. “But mostly, the domain was only $9.99,” he says.
The company started out developing a career guidance site, using big-data analytics to help people make career decisions. But as the site started pulling in job listings to show what was out there, beta users were asking for help with specific job searches. Levy and Smith realized they could refocus on helping users find jobs in the tech sector.
To do this, Beansprock “uses natural language processing to go out to the universe and find your best job out there,” Levy says. Its software pulls in job listings and combs through company information from sites like LinkedIn, AngelList, and Freebase. It’s able to understand what core skills are needed for a given job listing—JavaScript, Android, agile development—and it can handle different ways of describing skills and positions (“JS” for JavaScript, for example). It also tries to ascertain what markets a company operates in, how big it is, and so on.
Just as important, the software needs to understand who the job seeker is, too. The company’s algorithm takes as input from users things like their skills, experience, corporate culture, and tech-sector preferences (e-commerce vs. advertising, say). It also detects a user’s “level of position,” Levy says, so as to provide better job matches according to seniority and experience. The algorithm also learns from user feedback: with a Tinder-style mobile interface, a job seeker can swipe one way to approve a match and another way to reject it.
The matches are the result of a “knowledge base” that Beansprock has built up around some 10,000 job skills and how they relate to one another. “A lot of expert knowledge went into creating these taxonomies,” says Smith, the company’s CTO, who finished his PhD at the MIT Media Lab last year. “We can give the user insight into why they might choose the jobs.”
In other words, each job and each user is broken down into parts in software; each possible match is ranked based on all those parts; and the algorithm has human experts in the loop to help ensure the right mapping of users to jobs.
It’s an interesting application of artificial intelligence techniques, and it comes as a growing number of startups are creating virtual assistants, meeting schedulers, and smarter apps that employ natural language processing and machine learning. There are also “deep learning” approaches that, coupled with big-data and distributed computing techniques, are driving a lot of work in search, image recognition, language understanding, and decision making—see companies like MetaMind, Clarifai, Dato, Sentient Technologies, and of course Google and Facebook. Beansprock also fits in with an increasingly data-science approach to corporate hiring and skills assessment (see Cangrade, Smarterer, and others).
In the Boston area, where Beansprock is focusing first, Levy says the site pulls in “hundreds of new tech jobs” a day. “It’s tailor-made for a developer who’s somewhat content,” he says, but who might jump ship “if the right opportunity comes along.”
Beansprock isn’t making money yet, but it has plans to bring on customers from the hiring side of the marketplace who may be interested in providing better information about whom they’re looking to interview and their corporate culture. The startup’s strategy makes sense, considering how competitive the market is for hiring good engineers. But first, Beansprock has to get enough developers looking for jobs to use its site.
Levy, for one, is convinced his company has found the right niche. “Tech is a pretty big space, something we know well, and it’s buzzing. Especially if you’re a really talented developer—or anything IT, really, if you’re experienced,” Levy says. “You don’t need to settle.”
