Dr. Florian Sobieczky

Performing Data-Science tasks for Production Process-oriented Research.

In continuation of my work in the Design for Manufacturability group (DFM) of the austriamicrosystems AG, I was able to contribute an improvement of the statistical analysis of high dimensional electrical parameter datasets. I am delivering the full implementation of the code (R and MATLAB) as well as its benchmarking and performance comparison with formerly used codes.

For line production environments, data assessment is becoming a standard to the degree which allows "concurrent" fault-detection and localization procedures preventing production deficiencies of the final product at an early stage in the production line. I am developing a prediction method for this task using the full variety of modern machine learning algorithms.

The lecturer-ship for mathematics involved 7 yearly courses in the quarter system. It allowed plenty of time for research.

In the group for probability of the mathematics department of the University of Colorado at Boulder, I had one course to teach (Analysis) and was able to do research (Discrete Probability Theory).

Post-Doc position with a 4 hour lecturing commitment; Research (Discrete Probability Theory)

I could contribute to the development and implementation of a queueing model for a logistic supply chain network used by NovoNordisk. It''s applicability has been documented in a publication presented at ENBIS 2010 [1]. The model is an extension of the tandem queueing network used to cope with demand perturbations in logistic inventory chains [2]. [1]: http://www.enbis.org/activities/events/current/94_ENBIS_10_in_Antwerp/abstracts?abstractid=315 [2]: DOI: 10.1002/qre.1161

In my research project P18703 funded by the FWF (Austrian Science Fund), I was able to publish 3 papers related to random walks, percolation graphs, as well as structural theory of infinite graphs. At the end of this 3 year period (in the summer of 2009), I organized a workshop about random walks, spectra, and graphs in St. Kathrein (Styrian Alps).

At austriamicrosystems (production of analogue semi-conductors), I was able to use my statistical modelling skills to develop a quality control code for confidence regions corresponding to high dimensional data sets with a small number of samples; Methods included both non-parametric techniques, involving the `statistical (Tukey-) depth' ), as well as parametric models involving Bayesian approached to obtain density estimations.

During the time of completing my PhD, I had a teaching assistantship involving 4 weekly hours of teaching + preparatory work in Analysis, Linear Algebra, ODE, and Mathematical Physics.

Within a small image analysis company (`Algovision'), I was part of a team developing a compression code. My field was the segmentation problem, which I attacked with methods from probability theory (random walks). The product initiated the formulation my PhD thesis about estimating the number of connected components of percolation graphs.

Stochastics, Discrete Probability Theory, Percolation Theory, Random Walk Theory

Theoretical Solid State Physics