Swiss Post

Applying machine learning and statistics for the improvement of logistics in the whole of Switzerland.

Senior Data Scientist

Prognolite

Using statistics, machine learning, analytics and science to reduce food waste in the gastronomic sector. Building tools to make gastronomy easier.

Data Scientist and MLOps

Contract work:
Optimizing Switzerland-wide transaction intra-day liquidity and credit-risk, processing TB's of data.

Used technologies: 
Spark (Pyspark, Impala, Spark Scala), Jupyter Notebook, Tableau for visualizations for business. Bash and gitlab for CI/CD

Data Science Consultant

Knowledge Lab AG

Contract work:
As Data Scientist Consultant at Knowledge Lab I have been developing both internal and external projects.
Internally, I have been developing a Data Science management methodology in addition to researching and implementing several banking industry specific use cases, such as predicting customer churn, customer lifetime value,  fine-grained customer segmentation, data simulation and fraud detection. For all use cases, I developed a front end mock-up via Streamlit or R Shiny, so that I could refine the product together with our UX specialist for the end-user. In addition, I am responsible for scientific communication to stakeholders and the public, contributing to several texts and participating in a Statistical Programming Conference.
Externally, I developed a complex API layer for data querying and transformation for a client in R. Complementary, I provided data quality assurance tools and automated reports/documentation generation, deployed dashboards and the corresponding server software layers. All solutions were provided as installable packages via GitHub bundled with documentation generated from the source code and tutorials.

Main technologies: Nginx, MySQL (interfaced via R), R, Latex, Python (pipenv, pytest, Jupiter notebook, sklearn and similar packages), HDF5, R Shiny, Streamlit.

Contact work:
For a short project, SIX required a Data Scientist to build fault detection tools for subsequent error profiling of two core systems: the index calculation engines for the Swiss Business Index. A mathematical framework for comparing indices, generated by two different
systems, was formulated and implemented. This lead to the discovery of system wide deficiencies, not detectable by standard software engineering techniques.
Main technologies: SQL (interfaced via R), Scala (binding with core systems), R (data analysis and automated reporting).
• Designing data solutions for Swiss financial data services.
• Statistical modelling & dashboard reporting, applying data solutions to finance-related data-generation processes.

EPFL (École polytechnique fédérale de Lausanne)

R package development, to be found here: https://github.com/Seneketh/StanDDM

The package performs following:
-Prediction of human decisions via a multilevel Bayesian model in 'Stan'. Using Hybrid Monte Carlo for exploration of particularly  bothersome parameter spaces...

-Data Analysis: (Multivariate) linear/logistic mixed models of behavioural and EEG data in R

-Linking of Neuronal and Behavioural Data in the same fitting run.

-Building tools for model comparison in R (WAIC, LOO-CV, Lppd)

-Posterior simulations of model fits in R and Python (Pandas, Numpy, Scipy etc.)

-EEG (pre)processing in Matlab with EEG toolbox (PCI/PCA) and conduction of behavioural decision-making experiments

As a visiting Researcher at the Laboratory of Cognitive Neuroscience (Blanke Lab) I am building custom bayesian hierarchical (multilevel) models in the probabilistic programming framework 'Stan', of decision making and EEG/fMRI data to study the phenomenon of consciousness/metacognition for my masters thesis.

Computational Modelling/Machine Learning package developer in R

University Medical Center Groningen (UMCG)

-Applying machine learning/statistical models to  fMRI data in Matlab (Dynamic causal modelling) 

-Data Analysis: Multivariate linear mixed models of fMRI modeling results, clustering, multiple (bayesian) T-test, and bayesian factors in R

Dynamic causal modeling of fMRI recordings of verbal and auditory hallucination (AVH) episodes, stemming from 10 schizophrenia patients and 12 healthy subjects. The aim of the paper is to infer the underlying, causal neuronal systems of AVH in effective connectivity (EC) terms. 16 models of EC, exploring the hypothesis that the disconnection between cortical (STG and IFG) and subcortical (putamen and hippocampus) brain regions is responsible for AVH have been conceived. The model that fits the data best conserves all connections between brain regions. No effective connectivity differences have been found between healthy subjects and patients. Significant effects of AVH could be found on the EC from the putamen to the superior temporal gyrus. --- For additional material please visit this link: osf.io/tesw5

Because of legal reasons, the full Research Report and/or its presentation is available only on request. 
The project has been successfully executed. Awarded grade: 9.5

Machine Learning/Statistical Models of fMRI Data

Rijksuniversiteit Groningen

Mentor: Support in AI and HMC projects, coding and analysis. Guidance/coaching for international students.

Teaching Assistant for the courses following courses in the Bachelor Artificial Intelligence at RUG:

- "Research Methodology, Data Analytics and Communication (BSc Artificial Intelligence)
- "Basic Scientific Skills" 
- "Cognitive Psychology" 

8 Hours per week (2 lectives and 6 tutoring/grading) for two month each course

Teaching Assistant and Mentor in BSc and MSc Artificial Intelligence/HMC

Presence Switzerland, Federal Department of Foreign Affairs (FDFA)

- Liason between the Swiss Department of Migration and the Spanish Foreign Ministry for flight 
authorizations and administrative notifications.
 - Accreditation of Swiss diplomats
 - VAT declarations for Swiss diplomats
 - Translations German – Spanish / Spanish – German / German - English
- Provision of customer information by phone, e-mail and reception desk inquiries

Administrative Clerk

Universidad Autónoma de Madrid

My research experience in cognition and neuroscience focused mainly on basic EEG data analyses – preprocessing, defining trials and time-frequency analyses. This work was being performed under the supervision of Dr Capilla in the framework of her research in perception and attention, and using of the Fieldtrip toolbox in Matlab.

Research collaborator in Perception and Attention

I provide solutions which aid human decision-making in businesses through automatizing repetitive processes, reducing complex information, quantifying uncertainty, or offering guidance through predictive methods, combining science, programming and business know how.

Taplio

Marco Wirthlin's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Marco Wirthlin's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Marco Wirthlin's Linkedin Analytics