2 Comments

Summary:

Booz Allen Hamilton serves clients in areas such as national security, financial services and life sciences, and data science is an increasingly important part of the job. A VP in its Strategic Innovation Group talks about how he approaches hiring data scientists and analyzing clients’ data.

Booz Allen Hamilton is a consultant to some of the biggest companies and government agencies around, and it’s always hiring data scientists. What’s it’s looking for might surprise you, though: Unlike large web companies who often seek advanced degrees, mad computer science skills and — of course — plenty of experience in working with large data sets, Booz Allen Hamilton might well hire a curious musician.

Josh Sullivan, a vice president in the company’s Strategic Innovation Group, called into this week’s Structure Show and talked about what he looks for in a data scientists, as well as the utility of tools like Palantir and the shift in thinking that needs to happen for the intelligence agency to truly capitalize on the big data era. Here are the highlights, but you’ll want to listen to the entire show for all the context (download options below).

Download this episode

Subscribe in iTunes

The Structure Show RSS Feed

Don’t miss the point: Data science is not business intelligence

“A lot of people, they take their business intelligence users or their dashboard users and they say, ‘Well, now these are my data scientists,’ and they just kind of say, ‘I’m sure they’re doing data science work.’ I think they’re actually missing the point,” Sullivan said.

Traditional BI users or analysts already know the questions they want to ask, but they’re not forming hypotheses, testing them and then thinking of the next best question to ask the data. “They’re not doing ideation or accretive thinking,” he added.

Joshua Sullivan

Joshua Sullivan

Even software like intelligence-industry darling Palantir, which lets analysts parse through large systems of connected entities, isn’t data science as much as it is the product of a data scientist’s work. “We have a lot of people who use Palantir … that is a great tool for a domain expert,” Sullivan said. “… That’s something a domain user would use to consume the output, I think, of what a data science team would produce [working much closer to the data using statistical software and data-engineering techniques].”

Unicorns do exist — but good luck finding them

“Everyone’s out looking for the unicorn … the data scientist who can do everything,” Sullivan said. “… There are very few if any at all.”

Rather, Sullivan said he likes to hire a team whose members can play off each other like a jazz quartet. In fact, he has hired musicians before. He has hired a guy with a degree in forestry management, too. You don’t need to be an expert in machine learning and Python and Hadoop, because the chances are someone else is and they’ll carry that load until everyone else picks up some of that knowledge.

“It’s all about curiosity and then having some amount of skills and willingness to learn, I think,” he said.

Why work at Facebook? We have better data

Sullivan acknowledges that Booz Allen Hamilton does have to compete against companies like Google and Facebook for data science talent but, he added, the consulting does have some advantages over those more-glamorous gigs. Among them is the wide variety of data its clients in industries such as financial services, pharmaceuticals, oil and gas, international business and intelligence have, and the ability to put your discoveries to work across all those spaces.

“Some of those companies just have a big graph of people,” Sullivan said. “… For a hungry data scientist who really wants to dig in and understand the art of the possible, they need to have as much data and different varieties of data as you can give them.”

Do a little data sightseeing along the way

Sullivan’s group was helping a pharmaceutical-industry client get started with its data science efforts, including answering its primary question of how it can improve drug discovery. The team ingested all sorts of the company’s data — from supply chain info to sales data to handwritten research notes — and got to work trying to answer the question. Along the way, the team uncovered something that could let the company produce higher volumes of vaccines and drugs on which supply was currently short.

“One of the quick things we found along the road of ‘how do we speed drug discovery?’ was some of the substrates and some of the other products they were using to grow vaccines were actually not optimal,” Sullivan explained. “… This was several months ago and we’re already seeing millions and millions of dollars in additional yields for them to grow human vaccine. … They’re all about discovery of what’s in the data — very powerful.”

Thinking still needs to catch up with technology

Despite the great things that are possible with data science, though, Sullivan acknowledged that even purportedly omniscient industries such as those doing national security intelligence need to update their ways of thinking. They’re building technology and thinking about things like governance and privacy for the n-tier systems of 20 years ago, he said, but advances in distributed systems and cloud computing open up a whole new world of possibilities.

“We haven’t, I don’t think, updated a lot of the way we think,” Sullivan said. “… We’re trying to pioneer a whole new space and different way of thinking, and some the tools that we have just aren’t up to the job yet. But I think we’ll get there.”

Feature image courtesy of Thinkstock Images.

  1. Most data science people here and elsewhere cannot define data science and data scientist. They just end up confusing things up. In reality, these data science touting idiots are just users of some black box software and thats why they prefer ML.

    Share
  2. R. Robert Carlise Sunday, October 13, 2013

    Mr. Sullivan is spot on. Too bad those in the data analytics community don’t want to face that fact.

    Share

Comments have been disabled for this post