Curriculum Vitae
An intellectually curious problem-solver with 13+ years of experience across diverse industries, translating complex business problems into well-architected software & AI solutions, creating clarity from concept to code.
On the left you find my education and work experience, on the right my personal projects, books, and talks. The colored lanes in the middle are the threads running through it all; they grow thicker the more feeds into them. Click a thread's chip to highlight it, or any entry for details. Prefer the classic format? Download the PDF.
I'm a "get sh*t done (well)" person, who motivates empowered teams through a strong vision and clear priorities, while striving for operational excellence in an agile environment. I focus on user needs and make sure we're building the right thing.
Discovering patterns in complex datasets, crafting insightful data visualizations to guide decision-making, and building ML models that capture the true relationships between variables.
Tools: Jupyter, scikit-learn, PyTorch, pandas, streamlit, MatLab, R
I design software that's easy to understand, extend, and maintain, turning complex domains into clean, modular architectures.
Tools: Claude Code, Python, FastAPI, SQL (PostgreSQL, SQLite), Svelte, Kotlin (Compose Multiplatform), git, CI/CD, AWS, Docker, Kubernetes, Terraform, Argo Workflows, Grafana, Sentry
Proficient in professional writing and public speaking, enthusiastic about structuring information, clarifying complex concepts with diagrams, data storytelling, and meticulous editing.
Languages: German (native), English (fluent), French (basics), Spanish (basics)
Freelance Software Design Consultant
- taking custom software projects from the first conversation to running software, accelerated by AI coding agents
- mapping out processes together with the people doing the work to identify where software or automation (with or without AI) would pay off
- empowering teams to maintain and extend their software themselves
Business Analyst (eFTI EU-Project)Toll Collect (contract)
- devised a lightweight alternative to the eFTI platform software certification, incl. the implementation of automated test reporting
- represented Germany as an expert in the eFTI eDANOS certification working group
- created communication materials on eFTI, the new Platform Readiness Assessment, and our open-source strategy for internal and external stakeholders, incl. presentations to members of the European Commission
AI ConsultantWPS - Workplace Solutions
- built a RAG-based web app for generating tender documents
- implemented low-flow indicators for NIWIS in Python
- systematically evaluated AI coding agents for various refactoring tasks
- created a new training program "AI - Realizing Opportunities"
Fractional Tech Leadalcemy (freelance)
- planned and drove larger refactorings (creation of a monorepo; simplification of the ML and data ingestion customer configuration)
- built internal tools to improve processes, e.g., a forecast configurator and interactive feature explorer
- led several customer implementation projects
- initiated and led exchange sessions between the sales and data teams
- supported the team through two big reorganizations; hired and onboarded a data scientist and mentored a junior team member
Sabbatical
- wrote the Clarity-Driven Development book
- worked on personal projects & learned new skills (Vue.js, Kotlin incl. Compose Multiplatform)
- joined the Carbon13 Venture Builder Program (Berlin Cohort 3, Sept-Nov 2024) with the intention of founding a climate tech startup
Head of Data & Solutions Engineeringalcemy
- hired and led a team of three data scientists and engineers, responsible for customer integration and data-related customer requests
- created team vision & strategy and prioritized projects & objectives together with the customer success, product, and sales teams
- analyzed, documented, and optimized internal processes, e.g., reduced onboarding time for new customers from 50 to 25 days
- joined pre-sales discovery calls to surface integration requirements and translate product capabilities for non-technical audiences
- designed & built web apps empowering the customer success team to conduct recurring customer data analyses independently
Senior Customer-Facing Data Scientistalcemy
- analyzed laboratory data from our customers (cement plants), e.g., to identify irregularities in their production processes
- reduced time spent on recurring analyses by implementing configurable, reusable report templates
- simplified data integration process for new customers by planning and implementing refactorings of core product components, minimized required code files per customer to 6 from 17
Freelance Data Science Solutions Architect
- strategy & ideation workshops with department heads and product managers to identify potential AI use cases
- design, implementation, and evaluation of data science solutions tailored to my clients' needs (using Python)
- 1:1 coaching sessions and multi-day trainings on how (and when) to use machine learning techniques in practice (50+ trainings for audiences ranging from aspiring data scientists to C-level executives)
- clients included BASF (consulting + coaching for ~2 years) and TRUMPF (regular trainings since 2019)
Postdoc Visiting ScientistML Group
- developed continuously evolving word embeddings that account for meaning changes over time (published at ACL 2021)
Ph.D. (Dr. rer. nat.) Computer ScienceML Group, TU Berlin
- in the machine learning group of Prof. Dr. Klaus-Robert Müller
- funded by the Elsa Neumann scholarship from the universities of Berlin
- thesis: Similarity Encoder - A Neural Network Architecture for Learning Similarity Preserving Embeddings; developed a novel NN architecture to map high dimensional data into a low dimensional embedding space, where arbitrary pairwise relations between the data points are preserved
- supervised bachelor and master students
- graduated magna cum laude
Data Science ConsultantBASF
- predictive maintenance project in collaboration with TU Berlin - designed, implemented, and evaluated time series analysis models to predict the degradation of catalysts in chemical plants (patented and published in Computers and Chemical Engineering)
- authored the open source Python library autofeat for automated feature engineering and selection
- mentored other team members and responsible for code reviews
Machine Learning Team LeadSpectrm
- established the machine learning team and hired two ML engineers
- implemented a content recommendation API for newspaper articles to promote our clients' content (Python Flask app)
- developed a chatbot "AI" to respond to user messages automatically (using RiveScript)
Data Scientistidalab
- advanced analytics consulting projects and ML algorithm development in Python, e.g., an NLP backend for automatic content classification (razorfish) and style prediction for curated shopping (outfittery)
- presented results to clients
- project management
M.Sc. Computer ScienceTU Berlin
- focus: intelligent systems (machine learning, big data) & computational neuroscience (at the BCCN)
- thesis: Knowledge Extraction from Complex Biological Texts - A Machine Learning Approach
- graduated top of my class (1.0 on a scale from 1 (best) to 5)
Student Research AssistantML Group
- research on text classification, unsupervised learning (word2vec embeddings, dimensionality reduction), and information extraction
- EEG data analysis at the Berlin Brain-Computer Interface Lab - developed and efficiently implemented new algorithms in MatLab
- three peer-reviewed publications (journal and conferences); short-term research stay at UCLA
B.Sc. Cognitive ScienceUniversity Osnabrück
- interdisciplinary study program including courses in neurobiology, computer science, psychology, artificial intelligence, mathematics, computational linguistics, neuroinformatics, and philosophy; taught in English
- thesis: Comparing and Combining Multiple EEG Features in Motor Imagery BCI - A Large Scale Study
- graduated with distinction (1.1 on a scale from 1 (best) to 5)
Research InternMIT (Massachusetts Institute of Technology)
- at the McGovern Institute for Brain Research / Gabrieli Lab
- sponsored by a DAAD RISE scholarship
- analyzed fMRI data using Python (published in JAMA Psychiatry)
Student Research AssistantFraunhofer Institute for Chemical Technology
- worked independently, responsible for collection of infrared spectroscopy data
Abitur (secondary school)Fichte-Gymnasium Karlsruhe
- final mark 1.5 (on a scale from 1 (best) to 5)
- 11th grade as a year abroad in Missouri (USA)
Open Return
AI-powered trip planner I rebuilt from scratch by orchestrating AI coding agents. Co-founder.
Keynote DWX 2026: AI Rewrote
the Rules of What to Build
As agents take over tasks users used to do by hand, software products have to adapt to stay relevant, while GenAI opens the door to features that weren't possible before.
ArchGraph
Generates an interactive dependency graph of a codebase from a plain-text description of its modules, highlighting architectural layer violations.
Clarity-Driven Development
of Scientific Software
Empowers researchers to code with confidence and clarity, approaching software design from first principles with a product mindset (also available on Leanpub).
A Practitioner's Guide to
Machine Learning
Explains which use cases are best tackled with ML, gives an intuition for how the algorithms work, and shows how to arrive at a working solution while avoiding common pitfalls (incl. programming exercises).
evolvemb
Python library for continuously evolving word embeddings to examine word usage changes over time.
PyCon & PyData Berlin 2019:
Automated Feature
Engineering and Selection…
autofeat
Python library for automated feature engineering and selection (500+ stars on GitHub).
SimEc
Neural network architecture for learning similarity preserving embeddings (core of my PhD thesis).
Classify Me! Why?
Interactive web app explaining text classification decisions with Layer-wise Relevance Propagation.
textcatvis
Exploratory analysis of text datasets by visualizing relevant words of classes or clusters in word clouds.
PubVis
Web app to visually explore the landscape of published research and find similar papers.
ConEc
Extension of word2vec to create embeddings for out-of-vocabulary words and words with multiple meanings.
nlputils
Python library for analyzing text documents (TF-IDF features, similarity measures, kNN classification, t-SNE).
Publications
- Exploring Word Usage Change with Continuously Evolving Embeddings
- Forecasting Industrial Aging Processes with Machine Learning Methods [arXiv]
- Similarity Encoder - A Neural Network Architecture for Learning Similarity Preserving Embeddings (PhD thesis)
- The autofeat Python Library for Automatic Feature Engineering and Selection [arXiv]
- Automating the search for a patent's prior art with a full text similarity search
- Predicting Pairwise Relations with Neural Similarity Encoders
- The DALPHI annotation framework & how its pre-annotations can improve annotator efficiency
- Context encoders as a simple but powerful extension of word2vec
- "What is Relevant in a Text Document?": An Interpretable Machine Learning Approach
- Discovering topics in text datasets by visualizing relevant words
- Exploring text datasets by visualizing relevant words
- Interactive Exploration and Discovery of Scientific Publications with PubVis
- Explaining Predictions of Non-Linear Classifiers in NLP
- Robust Artifactual Independent Component Classification for BCI Practitioners
- Increasing the Spectral Signal-To-Noise Ratio of Common Spatial Patterns
- Predicting Treatment Response in Social Anxiety Disorder From Functional Magnetic Resonance Imaging [free version]
- Combining Multiple EEG Features in Motor Imagery BCI