Senior Machine Learning Scientist
EPFL (Swiss Federal Institute of Technology)
Geneva Area, Switzerland December 2016 – present
Conducted and supervised academic and industrial research on machine learning for analysis and control of large social and technological systems.
Particular areas of expertise:
– Machine learning for the Internet of Things.
– Social, mobile and telecommunication networks.
– Groups of autonomous robots.
– Development of mathematical and statistical methods for novel types of data and applications.
Principal Data Scientist / Researcher in Statistics and Machine Learning
Berlin, Germany – 2016
Research lead in Statistics, Machine Learning and Data Science for the European Union-wide Industrial Research Programme xD-Reflect: “Multidimensional reflectometry for industry”.
Among the participants, 12 research institutes from 10 countries worldwide and more than 30 industrial partners from 10 different industries, such as automotive, paper, printing, packaging, cosmetics, coatings, plastics and steel industries, and e-commerce.
The results obtained by teams under my guidance, are being used to set new fundamental visual quality standards and guidelines for manufacturers in the European Union. This is expected to lead to improvement of appearance of goods in dozens of industries, improving competitive position of European businesses with multi-trillion combined annual turnover.
– Led research on machine learning for computer graphics and reflectometry, perception-based material appearance modeling, realistic models for sophisticated visual effects; design of experiments, active learning.
– Advised scientists and programmers from 7 departments of 6 National Metrology Institutes and research Universities, in 6 countries within the European Union, in machine learning, statistics, applied mathematics.
– Led large-scale virtual experiments using parallel computing, designed sampling and active learning strategies for robots with complex sensors, composed research reports for governmental foundations.
– Developed a universal machine learning approach to manifold-valued visual appearance data, based on those goals that are common both for computer graphics as well as for industrial applications. This development is essential for such application areas as virtual reality and e-commerce.
– Served as an invited data science expert and consultant for Businesses in 5 countries worldwide, across a broad range of industries and topics.
Researcher in Statistics and Machine Learning
Murray Hill, New Jersey, USA –
Conducted research in the Statistics & Learning Research Department as well as in the Strategic Industries Research Department, Berlin, Germany, in parallel.
The main expertise can be broadly summarized as applied research for the Internet of Things.
– Led research on Support Vector Machines, time series analysis, and on machine learning and statistics applied to mobile communication networks.
– Led 5 industrial research projects, consulted on 12 business projects, established collaborations with academic institutions, participated in negotiations with external partners and in cross-departmental meetings, composed 3 annual performance reports supporting 2 departments, gained access to major data resources.
– Advised researchers, engineers and developers of Bell Labs’ Strategic Industries Research Department in machine learning, statistics, applied mathematics; shared experience in organizing interdisciplinary research and establishing external collaborations; identified research directions with the highest impact.
– Conducted research on machine learning and statistics applied to Smart Grid and renewable energy, traffic simulators for e-vehicles and urban planning.
– Reviewed and interviewed research scientists and interns.
Group Leader / Principal Investigator
Stuttgart Area, Germany –
Principal investigator, independent research group leader in the field of “Statistical learning theory for autonomous systems”.
– Proposed support vector machines of a new generation, having infinite or data-dependent output spaces, and an unknown number of classes. Proved universal consistency of these SVMs.
– Managed the research group budget, invited and hosted external visitors. Conducted reviews of applications, interviewed applicants.
– Hired and directed one postdoctoral researcher.
Stuttgart Area, Germany –
Joint employment at the Max Planck Institute for Biological Cybernetics and at the Max Planck Institute for Developmental Biology. Worked on machine learning theory, statistical and mathematical aspects of machine learning, and computational structural biology.
– Carried out research on statistics and machine learning for graphs and networks, such as biological and social networks and the Internet. Proposed a novel computationally feasible method for the analysis of properties of giant networks.
– Proposed a novel k-NN scan estimator for networks, and developed a theory establlishing the properties of the estimator for applications to large data sets.
– Developed novel unsupervised learning methods for the community detection problem and for finding communities in random graphs.
– Applied the newly developed methodology to data analysis in structural biology, in particular, to analysis of massive data sets of images from cryo-electron microscopy.
– Reviewer for a number of top conferences in machine learning.
Eindhoven, The Netherlands –
Carried out research on statistical and probabilistic image analysis, spatial statistics, statistical physics and statistics of networks.
– Proposed a groundbreaking method for the processing of noisy images. Developed the theory and implemented the method in R. Proposed algorithms are substantially quicker than competing methods, and have an extremely high accuracy.
– Applied the method to areas where massive amounts of data are collected and processed automatically and in real time, such as the insect vision modeling.
Cologne Area, Germany –
Conducted research at the Hausdorff Research Institute for Mathematics and at the Institute for Applied Mathematics. Worked in stochastic analysis with applications to Mathematical Finance and Physics, and in discrete probability theory, in particular on random walks, random graphs and MCMC theory.
– Proposed a new method for proving inequalities for stochastic integrals, leading to strong results in estimation of stochastic processes. Worked on application of these results to study solutions of Stochastic Differential Equations.
– Taught courses at the Institute for Applied Mathematics:
Mathematical Finance (SS 2008), with topics including arbitrage, assets, portfolios, derivatives, market models, utility, portfolio optimization, risk measures, value at risk, hedging.
PhD advisor: Prof. Axel Munk.
Teaching assistant: Asymptotic Statistics (WS 2005)
Business Intelligence Manager
St Petersburg, Russia
Founded in 1998, Falk eSolutions AG was a global provider of sophisticated ASP-solutions for online marketing. The core products, Falk AdSolution, Falk AdSolution FX and Falk MailSolution supported more than 300 vendors and website-operators as well as 50 agencies worldwide as an innovative AdServer-platform and a Rich Media management system, accompanied by an efficient e-mail-marketing system. Among the users were companies such as Ad Pepper Media, A & E Television Networks, Universal McCann and Mediacom. Falk eSolutions AG was later acquired by DoubleClick and eventually by Google.
I joined the company in 2004 as the Business Intelligence Department Head and was research and development lead in data mining, statisitcs, machine learning, optimization and algorithms applied to internet advertising.
Particular application areas included ads response prediction, contextual advertising, behavioral targeting, user modeling, log analysis, advertising metrics, A/B testing and real-time big data analytics.
Falk AdSolution became the third-largest ad management solution worldwide, serving over 30 Billion ad impressions per month.
Georg August University of Göttingen
Doctor of Philosophy (Ph.D.), Statistics
Thesis: “Data-driven goodness-of-fit tests”.
– Pioneered a general mathematical framework and developed a powerful method for constructing data-driven tests of statistical hypotheses for the case when observations are available only indirectly. The tests are incorporated with model selection rules. Proved consistency of these tests and model selection procedures.
– These results lead to groundbreaking applications in statistical inverse problems. Notably, I solved deconvolution density testing problems with both known and unknown error densities.
– Applications include statistical inverse problems arising in signal processing, imaging, economics, biostatistics; semi- and nonparametric statistics, including statistics for stochastic processes, time series analysis, Gaussian processes; model selection and variable selection; machine learning, in particular, for establishing universal consistency of SVMs.
– PhD advisor: Prof. Axel Munk.
Eindhoven University of Technology
Post doctorate, Statistics
Rheinische Friedrich-Wilhelms-Universität Bonn
Post doctorate, Probability Theory and Stochastic Analysis
Saint Petersburg State University
Master of Science (M.Sc.), Mathematics
Thesis: “On the generalized Goldbach Conjecture”
Native or bilingual proficiency
Native or bilingual proficiency
Professional working proficiency
- Elementary proficiency