All work
Demand forecasting for Sol Flower
A multivariate demand forecasting pipeline for Sol Flower, a US retail chain, trained on its point-of-sale history.
- Client
- Sol Flower
- Year
- 2026
- Role
- Head of AI at sense.ai
- Stack
- BigQuery
- Vertex AI
- Kubeflow
- TiDE
- ARIMA+
The problem
Sol Flower needed demand forecasts for each of its locations, built from years of point-of-sale history.
What I built
A multivariate time-series pipeline that models the historical sales data in BigQuery, trains deep-learning TiDE models against ARIMA+ baselines, and runs on Vertex AI and Kubeflow. Retraining and evaluation are automated for every location.