Partilhar via


Mathematical thinkers are key to our future: Microsoft is applying machine learning to all major applications

Editor’s Note: The following article was written by Michelle Feder and first appeared on the Microsoft JobsBlog

MVP Mentors can help high school and college students develop and improve their technical, professional and soft skills.  Fill out this short form to request your MVP Mentor. We apologize that the MVP Mentor Program does not provide professional consulting to adults candidates. If you are interested in learning more about the MVP Award Program please visit our site .

A report by the McKinsey Global Institute found that in the U.S. alone, by 2018, there will be 1.5 million jobs that require data analysis skills. As big data continues to grow, the industry needs people who have the quantitative chops to dig in, make sense of it, and make it actionable. When examined, mined and sorted, data can be winnowed into gold, yielding insights that can help guide how the concept for a product or service is shaped and engineered. This process of alchemy has a name: data science.

Mathematically minded people who have the talent to analyze and derive meaning from data, as well as those who know how to deploy techniques of machine learning, will find ample opportunity at Microsoft. Today, every major application at Microsoft is applying machine learning to enrich the customer experience and deliver new, intelligent capabilities. For scientists who’ve got algorithmic ability, this skill set can be the ticket to one of the hottest jobs on the market.

Whether you’re still in school, already in industry, or considering a shift, you may want more clarity about these careers. So let’s get back to basics: How do you define data science? And what exactly is machine learning?

For the straight talk, we spoke with industry pros as well as one industrious recruiter. Here, they share their views on why these trending professions are not just today’s buzzwords, but rather, careers that offer long-term growth at Microsoft.

ML Transition_1 (11)

Data Scientist Archana Ramesh has a doctorate from Arizona State in computer science, with a focus on data mining and machine learning, applied to cancer genomics. After earning her Ph.D., during which she had summer-interned at Google, Archana spent a year as a data scientist with a startup because she wanted to learn best practices—fast. In June 2014, she joined the newly formed Core Data Science team within Microsoft’s Operating Systems Group (OSG), the group developing Windows. It’s huge: The group includes Windows, Windows Phone, and Xbox.

The OSG Core Data Science team has an ambitious mission: Uncovering business insights to enrich the customer experience with Microsoft operating systems. The team brings technical and business expertise, plus engineering context, to enable data-driven decisions across Microsoft. This involves working with massive amounts of real-world data. “We make sure the engineers are all thinking in a way that’s based on data. We pick seed areas to focus on from start to finish, and we educate the engineers about data science,” Archana says. “Our goal is to apply data science and machine learning to the fullest.”

Asked to articulate the difference between data science and machine learning, Archana cites a teammate who sees machine learning as a two-part subset of data science. The first aspect is a question of: “Can we listen to the users to improve the product?”

For example, users of Internet Explorer can report problems they face when they’re surfing the web. To help resolve reports quickly, Archana applies a machine learning algorithm to collect customer comments. For developers in her group to use, Archana creates a graphic, which depicts the top problems. Machine learning applications make quick work of issue-spotting – which makes web browsing better.

The second kind of machine learning is a matter of “How do we use data science and machine learning to build a smarter product?” Prime example:  Cortana. In this case, “The machine learning is embedded in the product, to create a more personalized customer experience.”

Another Microsoft Data Scientist, Ioana Marginas, derives insights from telemetry-generated customer usage data about how people play games. For instance, she may observe that certain users play via touch screens rather than the mouse. She may note what time of day users are playing. What’s her process? “Most of my workday is spent in data cleaning and sampling,” she says. “After that, you can apply various algorithms to better understand the data.

ML Transition_2(5)

Ioana has a master’s degree in computer science, with specialization in information architecture, from Delft University of Technology in the Netherlands. She joined the Operating Systems Group in May after spending the previous eight years with IBM in Amsterdam. For Ioana, the transition to Microsoft was a logical and exciting career progression. She says, “I spent so much time in the Linux and mainframe world, that joining Microsoft and in particular, Windows, was the next natural step.”

Ioana interviewed with a number of companies to better understand the field in the U.S. market. She had multiple offers and decided on Microsoft for a few key reasons: She was drawn to the team’s entrepreneurial spirit and the nature of the work: It felt like joining a startup. “During the interview loop, the people I met showed passion and technical excellence. We were a newly created unit – like a startup within Microsoft — that was moving quickly. The team was flexible, diverse, fast, and focused on making a real impact in OSG.”

For Ioana, the chance to create something new was a pivotal decision point. Throughout her career, she has always worked in data. “I was excited about the opportunity to drive a real data culture in OSG and Microsoft. I had to grasp it.” Looking back, “The job was great when I first learned about it – and it is still great!”

Asked what she likes best about data science, Ioana smiles, chuckles, and says, “I’ve been in data science since I was born.” Her mother is a mathematician, her father an accountant. “They were not doing data science per se, but I was surrounded by people working with data.” Today, using tools and programs such as R, MATLAB, Python and Power BI, Ioana enjoys the “freedom of defining the question you want to tackle.” As a data scientist, Ioana says, “Although you do have a domain—an area of expertise—you can steer your projects in a way that allows you to make the biggest impact.”

ML Transition_3 (1)

Ram Shankar has one of the most inspired job titles at Microsoft. A member of the Azure Security Data Science Team, Ram calls himself a “Security Data Wrangler.”

A cloud cowboy?

Ram’s goal: Leverage the vast amount of data generated in Azure, to keep it secure. “I chose this moniker because here in the Wild West, every day is an adventure, and I need to wrangle the malicious hoodlums trying to attack the system. Only, instead of a double-barrel gun, I use statistical analysis and programming.” In his role as a security program manager, Ram uses machine learning because “you need intelligent security.”

Ram graduated from Carnegie Mellon University with a master’s in electrical and computer engineering, and pursued another masters in engineering management. He has been with Microsoft since 2012, when he joined the Office 365 Security Team as a security program manager. He says, “That team was one of the pioneers of using data science to improve security of the cloud.”

Today, Ram oversees data science projects his team does in collaboration with the company’s various research groups. On his own team’s security operations projects, he helps software engineers who work on data-driven applications test prototypes, and he provides guidance on constructing the right algorithms to ward off potential “attackers” automatically.

Ram digs in for answers. He explains, “There are many different data sources, for example various application logs, we use in Azure, and any new one remains foreign ground until we can deep-dive into it and explain what the data actually means: the syntax, the semantics, the merging patterns and outliers.”

Just how big is the impact of machine learning? Ram says: “Machine learning is now like programming: Everyone uses it, and at Microsoft there are so many opportunities for people from all disciplines.” For instance, two years ago, Ram interviewed Dan Mace, now a senior software engineer on the Active Directory Machine Learning team, who holds a doctorate in biology. “Dan had a strong emphasis on the computational aspect, where he used Bayesian modeling for studying cell expression. Today, his tools are widely used in security investigations.”

And just as practitioners can come from a wide variety of disciplines, machine learning technologies are being more widely applied in many areas, such as economics. Ram points to renowned researchers such as Microsoft’s Chief Economist, Preston McAfee, who used techniques from the field to redefine how auctions are done online.

“There is a lot of advice out there on how to become a data scientist, with the traditional emphasis on the computer science department,” Ram says. He welcomes wranglers from other fields of study to consider it.

ML Transition_4

Recruiter Amy Ala, who hired Ioana, agrees that many academic paths can lead to careers in data science. “Candidates need to have a solid background in computer science, modeling, statistics, analytics and math,” she says, but people coming from physics, astronomy, biomathematics, bioinformatics and related subjects are also a great fit.

Amy offers some ways a candidate can stand out: patents and publications, a Ph.D., strong communication skills and business acumen, plus the ability to pinpoint questions that will have the most value to the company. Whatever the practitioner’s original area, in quantitative fields, Archana says, “Data is data. Whether text data or numeric data, the core techniques and skills are similar.”

Archana says that in the “suite of skills” for a machine learning or data scientist, in addition to having access to one of the prototyping tools Ioana mentioned, it’s also helpful to know some variety of SQL as well as basic coding in languages such as C # (Sharp) and C ++, which can be useful when you need to build your own tools to solve some of the tough problems in this new frontier.

Archana is passionate about her chosen career. “I love dealing with data because every time you work with information, it presents new problems and new challenges,” she says. “These small nuances force you to think out of the box.”

And at Microsoft, one thing is for sure: As we build the next generation of machine learning platforms and powerful consumer applications, we’ll need data wranglers – and machine-teachers to translate data into useful information to make work and life ever better for customers.

Comments

  • Anonymous
    January 31, 2015
    good work...