Work in progress: machine learning x electricity markets
I spent two years making LLMs useful: a student-facing AI tutor in production (steacher.org, 15k interactions so far) and a few billion tokens burned writing code and crunching data.
Now I am pointing that toolbox at a harder target: probabilistic forecasting for power systems, in particular the imbalance a grid operator has to correct every 15 minutes when wind, sun and demand do not show up as planned. Less "what will happen", more "how wrong will we be, and how will the market react once we say it out loud".
Background: MSc in economics (ETH Zurich), 15+ years building software, a detour through marketing research at Coop, a few years of NLP.
Base:
- teaching computer science at HEVS university
- mountaineering
- father of 3