Case study

SunStreet energy price predictor

48h forecasts across 11 Swedish energy markets — ~50% revenue upside potential.

TensorFlow · Time series · GCS · Parquet

Jacob presenting the energy price predictor at AWS Stockholm

Presenting on stage at the AWS Stockholm hackathon

Overview

Built for SunStreet Energy AB at the AI Business Hackathon Grand Final (Kolomolo with AWS, Stockholm, February 2026). Specialised models deliver hourly 48-hour-ahead forecasts of spot and ancillary service prices so solar and battery operators know which market to sell into, hour by hour. Took 2nd place in a 40-hour sprint.

11 specialised models for Swedish spot and ancillary markets
Hourly resolution, 48 hours ahead
Leakage audits, incremental Parquet pipelines, weekly retraining on GCS
Potential ~50% revenue growth for solar and battery operators
2nd place — AI Business Hackathon Grand Final, AWS Stockholm