Ever since he can remember, Kriste Krstovski always wanted to be a scientist. Beginning in fifth grade in Macedonia, where he was born, he began competing in computer science contests. And he spent his boyhood summers taking programming classes, mastering Basic, Pascal, and C.

“Even as a boy I was curious and loved the challenge of studying new things,” says Krstovski, now an associate research scientist at the Data Science Institute, and also an adjunct assistant professor at the Columbia Business School.

Later, Krstovski studied electrical engineering at the University of New Hampshire. As a junior, he joined the Consolidated Advanced Technologies for Law Enforcement Laboratories (CATLab) as a research assistant. At the lab, he was a key contributor in the development of the Project54 system – a voice-based system that allows police officers to control all their in-vehicle electronic devices through speech commands. He stayed at UNH for a master’s in electrical engineering and as part of his master’s thesis he developed a handheld, voice-based, interface called Handheld Project54. The interface allows police officers to remotely control their in-vehicle electronic devices through speech commands from outside the vehicles. The interface enhances police officers’ safety and efficiency by providing them with the ability to perform drivers’ records queries using a single hand while outside of their vehicles. Project54 was a success: It’s now being used in the field by New Hampshire state police as well as by various law enforcement agencies throughout the nation.

After he earned his master’s, he worked as a staff scientist in the Speech and Language Processing Department at Raytheon BBN Technologies, a research and development company known for its pioneering work in artificial intelligence, especially speech recognition, natural language processing and computer networks. At BBN, he did research on DARPA projects that involved speech-to-speech translation as well as multilingual document analysis and translation. As part of the DARPA Transtac project, he developed one of the first handheld speech-to-speech translation systems. While doing research, he also spent two semesters studying at MIT as a special graduate student in the Electrical Engineering and Computer Science Department.

His next stop was at the University of Massachusetts Amherst for a doctorate in computer science. Under the direction of David Smith at the Center for Intelligent Information Retrieval, he did research in the fields of machine learning, natural language processing, information retrieval and machine translation. More specifically, for his doctoral thesis he researched efficient inference, search and evaluation approaches for latent variable models of text. While at UMass, he was awarded a predoctoral fellowship through which he spent the last three years of his doctoral studies conducting research at the Harvard-Smithsonian Center for Astrophysics. There, he continued his research on the efficient latent variable models of text for the SAO/NASA Astrophysics Data System (ADS). At the center, he worked with Michael J. Kurtz, an astrophysicist known for his research into the distribution of galaxies and for the creation of ADS.

After his PhD, he was a postdoctoral research scientist, working jointly with David Blei at Columbia and John Lafferty at Yale. While at Columbia, he was a member of the Blei Lab, where he researched unsupervised approaches for learning semantic representations of mathematical equations. He also led a team of undergraduate and graduate students in designing ArXivLab – a platform for developing and evaluating exploratory tools for the scientific literature.

Now, as an associate research scientist at DSI, Krstovski is continuing to do what he loves most: research. In his current research, he is exploring approaches in statistical machine learning, natural language processing and information retrieval. His aim is to develop tools that will help researchers and scientists comb through millions of articles and find the mathematical information they are interested in.

“Researchers need new tools to help them synthesize the ocean of information in their fields as well as related fields, so that they can make discoveries,” he says. “Profound discoveries often require scientists to find connections between their research and someone else’s. That’s why I love my current research. If I’m successful, I help other researchers make breakthroughs in their respective fields.”

Krstovski also enjoys teaching and interacting with students. This semester, he’s teaching a data-intensive course, “Computing for Business Research,” at the Columbia Business School. The graduate course is designed to teach business school students the essential concepts of computational methods and numerical algorithms – concepts that will help them conduct business research. In 2017, he developed and taught a natural language processing course at the Business School – a class designed to help Ph.D. students with their research.

“It’s exciting and challenging to teach data science concepts to students who’re studying accounting, marketing and finance,” says Krstovski. “Given the vast amounts of data across these fields, it’s imperative for them to know the computational methods we covered in class. And it’s fulfilling to me that I’m preparing them to succeed in their current research and in their future careers.”

— Robert Florida