August 26, 2013 11:00 AM PT
Each summer, Microsoft Research labs around the world assign mentors, set up extra workspaces, and welcome a new crop of interns. The internship program is a significant component of Microsoft Research’s partnership with the academic world.
It provides an opportunity for students to work at a corporate research facility, interact with world-renowned researchers, and network with a diverse set of fellow interns. In return, the labs gain fresh perspectives, forge stronger ties with universities—and enjoy the energy and excitement of talented young scientists.
The diversity of research disciplines at each lab means interns are sure to join projects that require their domain expertise. What makes the summer special, though, are the relationships the interns build with researchers and their fellow interns—a key goal of the program.
After completing an undergraduate degree at Koç University in Istanbul, Deniz Altinbüken is pursuing a Ph.D. in computer science at Cornell University, where she has been working on an open service that provides replication and synchronization support for large-scale distributed systems. Altinbüken’s background made her a perfect fit for the Distributed Systems Group at Microsoft Research Silicon Valley, where senior researcher Mihai Budiu has been her mentor.
“I got to join a project involving an infrastructure service for building storage systems in a cloud environment,” she says. “The project lets you specify exactly what you need and creates a matching storage system tailored to support data size, access frequency, read or update frequency, and other requirements. From your point of view, the system works like a normal storage system, but under the covers, it’s been customized on the fly to your exact specifications.”
Altinbüken has enjoyed tackling the tricky problems her team faced on this project. She especially likes how the modular nature of the system and the need for seamless integration between fundamentally different components have pushed the way she thinks about designing infrastructure software.
Perhaps a more important eye opener has been the work environment at Microsoft Research.
“A research career can involve a lot of bureaucracy,” she muses, “and that takes away from valuable research time. Sometimes, it can even limit your topics. This is not the case at Microsoft Research. It’s amazing to see how much freedom there is to let people work on fascinating ideas and enjoy their work. The culture here creates a great environment for innovative ideas.”
A Ph.D. candidate in the Learning Sciences and Technology Design program at Stanford University's School of Education, Matthews has been examining informal learning in online communities and how social-media practices alter our modes of acquiring and disseminating knowledge.
Working with her mentor, principal researcher Nancy Baym, and the Social Media Collective gave Matthews the opportunity to explore a different aspect of online communities. She has been able to study the evolving behavior and social norms that people must learn as they increasingly use online spaces for more aspects of their lives.
“The project asks what it means to behave professionally in a social-media world,” Matthews explains. “It studies how social media raises new and complex issues for the interactions between public figures and private citizens. I examined non-professional book reviewers and bloggers’ perception of their roles, as well as authors’ and agents' roles, and how those perceptions defined expectations for literary professionals’ behavior toward private-citizen reviewers.”
Matthews makes it clear how much she appreciates the helpful environment at the lab.
"These are top researchers working on multiple, fascinating projects,” she says, “but they still make time to help out interns any way they can. The thoughtfulness and care they take whenever they engage with me can't be stressed enough.
“To aspiring new interns, my advice is: Talk to people! There's a wealth of resources to be tapped from both permanent and visiting scholars. Everyone is so welcoming that you shouldn't be afraid to step out of your comfort zone or group to see what knowledge you can glean from them all.”
Halley Profita’s Ph.D. work at the University of Colorado Boulder is oriented around human-computer interaction (HCI) for health/assistive applications. She has a special interest in smart fabrics that use distributed computation to explore novel sensing and interface designs in garments. When it came to her summer project at Microsoft Research Redmond, though, program manager Donald Brinkman of Microsoft Research Connections challenged Profita with a unique opportunity that also used her Master of Industrial Design degree from the Georgia Institute of Technology.
The opportunity? To design interactive games and installations for the newly renovated Buildings 5 and 6 on the Microsoft campus.
“The goal was to breathe life into the workspace,” Profita explains. “We wanted to explore ways of integrating technology further into the workplace in order to promote company culture and encourage interconnectedness and general well-being. This meant addressing multiple factors such as HCI design, maintenance of personal health, gaming and engagement elements, workplace and life-space balance and support, and intellectual imagination.”
Profita used in-house technology, such as Kinect and .NET Gadgeteer, to create engaging, interactive installations for the buildings that complemented the office environment while offering an added health component. The result? A water cooler that flirts to garner increased usage and a wall unit that teaches calming tai chi moves.
Meanwhile, her own work environment at the lab has been a highlight of the summer.
“Microsoft Research is a place where interns are encouraged to explore far-reaching questions and apply out-of-the-box problem solving,” she says. “It’s been an invaluable experience. I’ve been able to interact and learn from experts in the HCI field, meet people from various product groups, and learn about the full gamut of initiatives at Microsoft. The opportunity to leverage my creative background to address pertinent research issues has strengthened my desire to pursue a research career.”
University of Cambridge student Abigail See has plans to pursue a Ph.D. in computer science after she completes her master’s program in mathematics. See always has been interested in game theory, and during her internship this year at Microsoft Research Cambridge, mentors Yoram Bachrach and Pushmeet Kohli, researchers with the Machine Learning and Perception group, gave her a chance to dive right in.
“We’re investigating a variation of ‘weighted-voting games’ in cooperative game theory,” See says. “A weighted-voting game models a system where agents, each in possession of a number of votes, can form a ‘winning coalition’ if that coalition holds enough votes. An example would be the formation of a coalition government from political parties that each has a certain number of elected representatives. Most models of weighted-voting games assume that all coalitions are possible—meaning that all parties are willing to work together. We are studying a model in which some parties are incompatible—just as in real life!”
“Perhaps the most challenging aspect of that internship for me was working on a large coding project with several other people,” See recalls. “Now, I’ve worked on two completely different projects and had the opportunity to learn from world-renowned scientists. Both experiences have been amazing introductions to the world of research. I love the vibrant atmosphere, especially the frequent lectures by visiting researchers, which are a good way to hear what’s going on in other areas.
“A Microsoft Research internship is a unique experience. You get an inside look at the world of research and the world of a technology giant.”
“This is one of the most exciting projects I’ve ever worked on,” says Anna Choromanska. “It gives me the opportunity to get very deep into several different machine-learning techniques—and then borrow from those techniques and invent to create a completely new machine-learning tool with guaranteed good performance.”
Choromanska is a Ph.D. candidate in the Department of Electrical Engineering at Columbia University. Fascinated by machine learning, her main Ph.D. project is about splitting complicated optimization problems into much simpler sub-problems.
At Microsoft Research New York City, Choromanska works under the guidance of mentor John Langford, senior researcher, and postdoc researcher Alekh Agarwal. The project that has caught her imagination provides faster access to massive data sets through accurate prediction of data-set labels.
“Massive data sets are everywhere—think of biological data or web data,” she explains. “An example would be YouTube videos, where each video is an instance in the data set, and labels are their categories: news, music, sports, or cats—any one of a huge number of labels. The core challenge is to design algorithms that have been trained on labeled data sets to predict the labels of instances in huge unlabeled data sets and to do it very quickly and accurately.”
Her internship at Microsoft Research has given Choromanska plenty of challenges and hard work, but she already knows she will miss the lab when her internship is over.
“The culture of intellectual freedom and intensive work is compelling,” she says. “I was really impressed that interns have easy access to all the researchers. There is an open and friendly atmosphere. It’s truly amazing to sit in front of a whiteboard with the most famous researchers in your field and derive theorems with them. It’s like being part of the dream team.”
Peiran Ren first came to Microsoft Research Asia in 2008 as a graduate student from Tsinghua University. Ren is part of a joint program between the Institute of Advanced Studies at Tsinghua University (IASTU) and Microsoft Research Asia, and his studies focus on realistic rendering of computer images, a topic he explores as part of the Internet Graphics Group, under the supervision of adviser Baining Guo and mentors Jiaping Wang and Xin Tong Lately, Ren has made a name for himself with light.
“Simulating light transmission in a photo-realistic manner is very complex,” he says. “Traditional techniques for computing full lighting effects—complex light interaction among many objects—can require minutes or even hours to render a single frame of image. In our latest project, we bring in machine-learning techniques as the key solution for realistic rendering. This allows full lighting effects to render in real time. This is really important for film or gaming applications.”
This unique approach demonstrated the effectiveness of combining computer graphics with machine-learning methods to solve a traditionally difficult graphics problem.
“It was an almost entirely new approach,” Ren recalls. “We didn’t have any direct references to available literature, and at the start, we didn’t even know whether the problem could be solved through machine learning.”
The resulting paper, Global Illumination with Radiance Regression Functions, which Ren co-authored with Wang, Minmin Gong, Stephen Lin, Tong, and Guo, caused a sensation when presented during SIGGRAPH 2013.
For Ren, the innovation would not have been possible without support from Microsoft Research and IASTU.
“IASTU has been not only a place to study, but also a place that is pure academic paradise,” he says, “where we can enjoy both study and research. At Microsoft Research, everyone is always very positive and approachable. I have learned so much from discussions and working together with Baining, Xin, and Jiaping. It gave me the confidence to explore a completely new research topic, even though I wasn’t sure if the approach would deliver results.”
All in all, the intern experience could be summarized by Altinbüken.
“The environment,” she says, “motivates you to get better at what you do, in ways you could never have imagined.”