Washington, District of Columbia, United States
Konstro entwickelt Software für schnellere Baudokumentation mit KI + Sprachfunktion -> www.konstro.de
Built an investment pipeline and network for a new VC-branch from scratch. Sourced, evaluated, and invested in Biotech & AI technology startups. Developed LLM-based tools to streamline the startup discovery and due diligence process.
Built data pipelines for large datasets, ingesting messy data from various sources (SQL, Spark Cluster, Standalone CSV-files). Analysed historical & simulated (counterfactual) datasets to assess the impact of construction scenarios on KPIs. Developed an inhouse Python library to streamline data analysis. The package was adopted across multiple teams and saved Data Scientists significant time. Established clean code principles in my team (decorators, typing, docs, formatting, testing). Held workshops about technical topics: pair programming using mob.sh, Window Functions in SQL. Tech Stack: Python, SQL, Docker, Jenkins, Git & GitLab, Jira, PySpark, Pandas, Polars, Psycopg2, Streamlit, SckitLearn and more
My research included a deeper look into novel approaches for valuing a customer base by dissecting it into cohorts of customers. I attempted to adopt so-called age-period-cohort models, which are statistical models popular in epidemiology, to common marketing questions. For example, I used those models to analyse and forecast customer retention rates and revenue retention rates. Besides research, I developed a fairly sophisticated R Shiny dashboard app that illustrates the machinery behind a customer-based corporate valuation. In particular, the app illustrates how a customer base can be dissected into customer cohorts. The dissection into customer cohorts allows us to analyze the evolution of a marketing metric (e.g., customer retention) across time, across cohorts, and across the cohorts' lifecycles.
Assisted research in Marketing Data Science. Developed an R Shiny application to conduct customer-based corporate valuations, which was used for academic and educational purposes.
Held tutorial lectures for Discrete Modeling, a course that teaches fundamental mathematical concepts required to translate real-world problems into mathematical language and, ultimately, code. I lectured in front of 20 students and reviewed my students' weekly assignments. I successfully faced the challenge of deciphering handwritten equations of first-year computer science students. Nothing scares me anymore.