About Gresp
Real-time, spatially aware regional forecasting, built on graph neural network research.
What we do
Gresp turns regional data into short-term forecasts. Instead of treating each region as an isolated time series, our temporal graph neural network (built on MPNN_LSTM) models how demand or risk actually moves between connected regions — so a forecast for one area accounts for what's happening in the places people, patients, or goods travel to and from. We started with epidemic outbreak forecasting and have since applied the same engine to pharmacy prescribing demand and taxi mobility demand, with retail replenishment and energy-grid demand also tested.
Who it's for
Public health teams, pharmacy and retail operators, fleet planners, and other operational teams who need faster, more spatially aware signal on where demand or risk is heading next — not just whether it's rising.
How it works
- Upload — bring your own regional data and a connectivity graph, or try it instantly on one of our live demos (epidemic surveillance, pharmacy demand, or taxi demand).
- Train & predict — the model learns the spatial and temporal patterns in your data and forecasts values at the horizon you choose.
- Download & act — get region-by-region predictions you can act on, exportable for your own reporting.
Built by
Afraz Khan
Co-Founder — Modelling & Platform
- Senior data scientist working in provincial public health surveillance
- Builds and deploys the forecasting models in production
- MEng Electrical & Electronic Engineering, Imperial College London
- Master of Finance, University of British Columbia
- Based in Vancouver
Vanessa Fanyu Xiu
Co-Founder — Product & Operations
- Trained as a mathematician, with 5 years' experience in population health forecasting in research and industry
- Cloud integration and UI/UX
- MSc Epidemiology, McGill University
- BSc Mathematics, McGill University
- Based in Ottawa