Georg Heimel


go to CV






Physics is the only profession in which prophecy is not only accurate but routine.

Neil deGrasse Tyson
© 2016 by GH: Ostsee-Impressionen

curriculum vitae

  present interest

Machine Learning | python
  • Probabilistic modeling

  • Graphical models

  • Deep learning

  • Bayesian decision theory

  skills

Data Science | machine learning
  • Data cleaning & visualization (pandas, matplotlib, seaborn, folium)

  • Statistical hypothesis testing (scipy.stats, statsmodels)

  • Clustering, regression, decision trees, SVMs, cross-validation, etc. (scikit-learn, RankLib)

  • Recommendation engines (collaborative filtering, truncated SVD)

  • Convolutional neural networks (Keras)

  • Probabilistic modeling (pymc, pyflux)

  • Time-series analysis (pyflux, arch)

Technology | tools
  • NoSQL and SQL databases (Elasticsearch, Yandex Clickhouse, PostgreSQL, MySQL)

  • Elastic stack (Filebeat, Logstash, Elasticsearch)

  • Apache Kafka, Strimzi.io, Apache Spark (structured streaming)

  • Docker, docker-compose, OpenShift (OKD)

  • Amazon AWS

  • Monitoring (Prometheus, Grafana), GitLab CI/CD

Coding | test-driven development
  • python, scala, ruby

  recent projects

Road recognition in aerial imagery
  • With the goals of quantifying the road-network coverage in OpenStreetMaps and improving it, convolutional neural networks (CNNs) were trained to identify roads in aerial imagery. A python package was coded on top of Keras/Tensorflow, integrating the communication with mapping services, image pre-processing, training and validation of CNNs, as well as predictions on trained models.

Custom data platform in lambda-architecture
  • Analytic data, logged to local disk in the VMs of backend services, was read with Filebeat and streamed to Logstash, where each JSON was validated by a custom plugin. Clean data was then passed on to Kafka, where Spark (structured streaming) was used to enrich it with geo-spatial information in real time. Finally, Yandex Clickhouse served as a data-lake with Apache Superset providing self-service access.

Ranking of points-of-interest full-text search results
  • With POI data stored in Elasticsearch, the aim was to improve the ranking of location search results displayed in mobile clients by machine-learning from customer app-usage. A python package was coded from scratch to explore and define features, to handle communication with Elasticsearch and its RankLib plugin, to upload trained models and search templates, and to graphically/statistically compare models.

Real-time supply & demand capture
  • To visualize and query discrete events as supply and demand that varies smoothly and continuously as a function of space and time, a computationally efficient algorithm for density estimation based on orthogonal polynomials was formulated and implemented in python, using process-based parallelization.

  career history

2017 - present | Sparks42 | senior data scientist Sparks42
  • Consulting a major ride-hailing venture on data science and data infrastructure

  • Design, implementation, deployment, tuning, and maintenance of lambda-architecture data platform

  • Real-time streaming data enrichment with geo-spatial information

  • Convolutional neural networks for road-network extraction from aerial imagery

  • Interactive framework for machine-learned ranking POI full-text search results

  • Highly parallel orthogonal-polynomial density estimation for real-time supply/demand analytics

  • Hiring and recruitment

2008 - 2017 | Humboldt-Universität zu Berlin | group lead HUB
  • Principal investigtor in research projects on computational materials modeling

  • Scientific computing on own HPC resources

  • Budget responsability for independently raised third-party funding (ca. € 1 Mio.)

  • Personnel responsability and supervision of doctoral theses

  • Consulting industrial partners

  • Managerial duties (collaborative research center, hiring committees)

  • Lectures at the Department of Physics (amongst others, on Computational Physics)

2007 - 2008 | Massachusetts Institute of Technology | postdoc MIT
  • Design of elementary building blocks for molecular electronics

  • Code development (fortran) and optimization (blas/lapack)

2004 - 2007 | Georgia Institute of Technology, USA | postdoc GaTech
  • Functionalization and optimization of materials and interfaces for organic solar cells

  • Organization and scripting of multi-tiered copmuter-simulation processes

  • Solving mixture models on spectral data from laboratory experiments

  • Prototyping quantum-mechanical models in python

2000 - 2004 | Technische Universität Graz | research assistant TUG
  • Quantum-chemical simulations of molecular semiconductors

  • Eigenmode analysis of molecular vibrations in Mathematica

  • Numerical solution of non-linear differential equations

  • Data retrieval with sliding-window Fourier transformation

  • Regression analysis of X-ray diffraction patterns

  education

2000 - 2003 | Dr. techn. (Physics) | Technische Universität Graz, Austria TUG
  • PhD thesis: Structure and Optical Response of Conjugated Molecules

  • graduation with honors

1993 - 2000 | Dipl. Ing. (Physics) | Technische Universität Graz, Austria TUG
  • MSc thesis: High Pressure Studies on the Structure and Optical Properties of Poly(para-Phenylene) Based Systems

  • graduation with honors

  • ERASMUS exchange in Grenoble, France

1985 - 1993 | Matura | Carneri Gymnasium Graz, Austria BG
  • graduation with honors

  publications

2000 - present | 90+ publications in peer-reviewed scientific journals
  • Thomson-Reuters ISI h-factor: 33

  • some highlights

    link Nature Communications, 6, 8560 (2015)

    link Science Advances, 1, e1501127 (2015)

    link Nature Communications, 5, 4171 (2014)

    link Nature Chemistry, 5, 187 (2013)

    link Nature Materials, 7, 362 (2008)

    link ...

  awards

Nov. 2012 | Junior Award for Good Teaching PHYS
  • Physics student's representative body at Humboldt-Universität zu Berlin

Nov. 2007 | Special Research Prize for Nanotechnology and Nanoscience Stmk
  • Category fundamental science

  • State of Styria, Austria

March 2006 | Marie-Curie Fellowship
  • European Commission

Sept. 2004 | Erwin-Schrödinger Fellowship FWF
  • Austrian Science Foundation (FWF)

  activities

1996 - present | Reviewer | publishing and funding
  • For scientific journals such as Nature Nanotechnolgy, Nature Communications, Physical Review Letters, or Advanced Materials

  • On conference such as Optical Probes, International Conference on Synthetic Metals, or PyData Berlin

  • For funding agencies, e.g., Deutsche Forschungsgemeinschaft (DFG) or Department of Energy (USA)

April 2015 | Guest Editor | scientific journal AFM
  • Special Issue 13 of Advanced Functional Materials, vol. 25 (2015)

May 2014 | Organizer | symposium international conference E-MRS
  • Symposium O at the Spring Meeting of the European Materials Research Society (E-MRS) in Lille, France

May 2013 | Panel Member | content and scope
  • Symposium J at the Spring Meeting of the European Materials Research Society (E-MRS) in Strasbourg, France

2000 - present | Speaker | conferences and seminars
  • 38 talks at international scientific conferences such as the ICSM, MRS, ECME among others in China, USA, Japan, Italy or Switzerland

  • 20 invited guest lectures, among others at the University of Groningen (Netherlands) or the Weizmann Institute of Science (Israel)

1999 - present | Visiting Scientist | short-term research stays
  • link University of Missouri, Columbia (MO), USA

  • link University of Arizona, Tucson (MO), USA

  • link University of Wisconsin, Madison (WI), USA

You have questions?


contact me ...





last update 10/2016