Thorsten Kranz
Passion for data.
Head of AI Solutions at DHL Group, physicist and neuroscientist by training. Here I write about generative and agentic AI, data science and software development, and collect my talks and side projects.

Find me online
Articles and talks
2026
Rocket Catch: Land a Super Heavy on the Tower
A browser flight sim of the Starship chopstick catch, with real masses and thrust, four tolerances to hit at once, and an autopilot that misses a third of the time.
Mock Mirror: Using AI for my own News Website
How I turned an AI journalism experiment into a fully autonomous news website that researches, writes and publishes every day.
Loop Velocity as Key Differentiator in AI Coding Systems
Why the speed of the write-run-fix loop, not raw model intelligence, will decide which AI coding agents win in 2026.
2025
The Chris Rod Max Show: Why Most Enterprises Fail at AI
A conversation about why AI adoption stalls in large companies, from data quality and build-vs-buy decisions to the future of jobs and AI literacy.
CUT 2025: Age of AI
My keynote at the 12. Cloud Unternehmertag: how DHL uses AI today, five steps towards AGI, and what this means for your business.
2024
Clean Code - What is it, why it matters
What Clean Code means in practice (readable, modular, continuously refactored code) and why it pays off for every team.
2023
Human-Centric Parcel Delivery at Deutsche Post with Operations Research and Machine Learning
Our Wagner Prize finalist paper on how Deutsche Post plans courier routes by learning from experienced couriers, enabling accurate delivery time windows.
Data Culture Podcast - ChatGPT at DHL
My second visit to the Data Culture Podcast: generative AI at DHL, from strategy and security to practical use cases in logistics.
PyCon DE & PyData Berlin 2023: Delivering AI at Scale
Behind the scenes of Deutsche Post DHL's journey to a data-driven company, from large-scale use cases and tooling to a time series forecasting library.
An overview of Generative AI
A short tour of generative AI: language models, image, speech and video generation, and the risks that come with them.
2022
5 Things I hate about Python
Five things that still annoy me about Python after 15 years, from the GIL to packaging, and what I use instead.
Data Culture Podcast - Data Literacy for 600,000 Employees
How to build data literacy in a company with 600,000 employees, plus data governance and a data and analytics organizational model at Deutsche Post.
The Erium Podcast : Scaling AI in Logistics
Talking with Timo Steininger about scaling data science at DHL: our Homebase model, delivery route optimization and building a data culture.
2021
Overview of Machine Learning areas
An overview of supervised, unsupervised and reinforcement learning, and the kinds of problems each of them solves.
Chasing unicorns, or: profiles of Data Scientists
Why the search for the full-stack Data Scientist fails, and how data engineers, data scientists and ML engineers cover the job together.
PyLocator: Interactively locate EEG/MEG electrodes in MRI Scans
PyLocator, my open-source Python tool for locating intracranial electrodes in MRI scans, and the research problem it solves.
Learning Data Science online
A guide to learning Data Science online: introductory talks, the courses worth your time, places to practice, and learning paths by background.
How to form a Data Science team
Getting the skill mix right, choosing an organizational setup, and building a culture that rewards the team over lone wolves.
















