Melanie F. Pradier joins me via Skype. It’s mid-afternoon in Madrid, the beginning of the work day in Cambridge. Pradier’s enthusiasm – her palpable energy and vivacity when talking about her research – is more invigorating than my cup of coffee. After a few minutes of technical difficulty, we’re off! I’m immediately struck by the breadth of Pradier’s experience. She has studied a wide range of disciplines in an even wider range of locations, making her just the sort of global researcher CRCS aims to attract.
As an undergraduate student at the Technical University of Madrid (UPM), Spain, Pradier studied telecommunication engineering with a major in digital signal processing. At that point, she was interested in mathematically-oriented fields related to data-mining. This interest led her to seek out a Master’s degree in information technology in Stuttgart, Germany, followed by two industrial research internships at Sony, the first in Germany and the second in Japan (Pradier has an abiding interest in Japanese culture and has even taught herself the language). Her experience in the industry made her realize the massive potential of machine learning and of probabilistic modeling in particular. This realization led her back to Spain, where she began pursuing a PhD at the Marie Curie ITN “Machine-Learning for Personalized Medicine” European Initiative. There, she acquired not only technical expertise but also bio-clinical knowledge. She attended summer school in the field of personalized medicine, made two stays at hospital research centers, and conducted joint work with “Roche,” a pharmaceutical company.
Unsurprisingly for a CRCS postdoc, the highlight of Pradier’s research has been her contribution to multi-disciplinary projects that have a direct impact on society. During her internship at Sony, she and her Japanese colleagues acquired a double patent for an adaptive learning technology intended for application to online education. Her work on personalized medicine projects contributed to the development of general tools for genetic association and biomarker discovery. These tools have the potential to directly impact and improve the health outcome of patients around the world. Pradier has found that these multi-disciplinary projects present their own challenges, as they require experts from disparate fields to communicate effectively with one another despite operating with vastly different technical vocabularies. That said, the experience has been a rewarding one for Pradier and has led her down the winding path to CRCS.
Pradier’s work now focuses on probabilistic graphical models and approximate inference algorithms for data exploration, including Bayesian non-parametric methods, scalable MCMC or variational inference, modeling of structured data, and statistical computation approaches. She is interested in both methodological foundations and real-world applications. Thus far, she has designed specific models for data exploration to answer questions in fields such as social science, cancer research, and economy. The question that drives her is how we might improve the process of data exploration, both in general (i.e. encoding prior knowledge into the model, making inference more efficient, and automating the step of making a model more interpretable) and in particular (i.e. answering specific relevant questions in diverse research areas).
We are only beginning to understand the extent to which data holds crucial information for addressing our society’s most pressing questions: What are the underlying mechanisms of cancer? What factors make some countries wealthier than others? Most of these questions cannot be stated as well-defined, discrete problems. They require multidisciplinary research efforts involving easily interpretable models and rigorous data exploratory analyses. Pradier’s goal is to turn data into useful knowledge in order to answer these urgent societal questions. To this end, she works regularly with experts in a number of fields. Thus far, she has had the pleasure of conducting research with oncologists, biologists, economists, computer scientists, and chemists. She stresses that these collaborations are two-way communication channels. Experts provide model validation and valuable insight about the assumptions appropriate to their respective fields while Pradier provides the statistical tools and models necessary to extract useful information from the data. In so doing, she expresses one of CRCS’ core values: multi-disciplinary collaboration is crucial if computational science is to have a positive impact on society.
During her upcoming postdoctoral fellowship at CRCS, Pradier intends to extend these collaborations by connecting with top experts in a number of disciplines, including computer scientists, physicians, and economists. Her plan is to develop general methodological tools as well as to work on specific multi-disciplinary projects. She will be part of Professor Finale Doshi-Velez’s research group, where she is especially excited to begin collaborating with researchers in areas such as healthcare and human-computer interface perspectives.
Pradier’s interests are not confined to her scientific research. When she is not hard at work, she enjoys jogging, hiking in the mountains, and ice skating. She enjoys learning about Japanese culture and is a competitive player of “Go,” otherwise known as Japanese chess. We at CRCS are thrilled to welcome Melanie F. Pradier to our community. You will find her delivering a CRCS seminar on November 20th!