Narendra Mukherjee
I am a Senior Machine Learning Scientist working at the Booking.com headquarters in Amsterdam in the Netherlands. I am broadly interested in Bayesian interpretations of machine learning/statistics, as well as software solutions to better productionize data and ML products. At Booking, I work primarily on ML models of multi-touch attribution and how these can have impact on marketing incrementality measurement and bidding optimization.
Previously, I have worked as a Research Scientist at Philips Research in Eindhoven in the Netherlands and as a a Machine Learning Scientist at Tripadvisor in Boston in the US. In these roles, I worked on a variety of ML topics ranging from recommendation systems and NLP to marketing budget optimization.
Prior to that, I earned my PhD working with Prof. Donald Katz in the Neuroscience graduate program at Brandeis University where I was also a Howard Hughes Medical Institute (HHMI) Predoctoral Fellow . My PhD work lies at the confluence of neuroscience, dynamical systems, machine learning and the evolutionary basis of animal behavior. In the Katz lab, I studied how the processing of tastes in the brain leads to consumption or rejection decisions. For this work, I performed chronic electrophysiological recordings from large multi-electrode arrays implanted in rodents actively engaged in consumption tasks. I built probabilistic (Bayesian) graphical models to understand the neural population-level firing patterns obtained from such large-scale recordings; these statistical models, together with precisely-timed optogenetic perturbations of neural activity, have revealed a unique temporally-specific role for sensory cortex in taste-related decision making. To know more about this work, please look at my PhD research.
I am deeply committed to making research and machine learning tools more accessible through open-source contributions. During my PhD, the centerpiece of these efforts was the development of a flexible open-source electrophysiology hardware and software system that can combine optical fibers for optogenetics with long-term recordings from over 1000 electrodes in awake rodents at a fraction of the cost of commercially-available solutions. Much of this work was supported by the Brandeis University Maker Lab and cloud computing resources on the Jetstream supercomputer of the NSF.
Aside from work, I am also passionate about cycling and how bicycles are key to sustainable cities of the future. Cycling has been a joy for me right from the time I grew up in India, and that hobby developed further in the years I spent cycling in the New England area in the US. However, I also started to realize how much of a center stage the automobile takes in modern cities (esp. in the US), and how that is an unsustainable, dangerous and unhealthy situation. Since moving to the Netherlands, I am learning more from the sustainable and equitable Dutch city planning expertise which puts the bicycle in a central position, and I hope to be able to write more on that in the future. Till then, browse the NotJustBikes YouTube channel to learn about the subject.
For more information, read my CV, look at my PhD research, browse my publication list or explore my projects on Github.