Best-in-Class Demand forecasting. 


Built on top of a suite of machine learning and traditional time series methods, our AutoML methods deliver the best model for every single product across your product range.

Built-in Data

Weather data and other data are built in, meaning you can quickly and easily understand the impact these factors are having on your sales.

You simply need to provide longitude and latitude of your outlets to understand how the weather is affecting sales, this can then be used in conjunction with current forecasts to help with the short-term forecasting.

Low-Velocity Forecasting

Specifically designed models deal with SKUs that don’t sell at a high frequency, but are still critical to forecast correctly.

Where other forecasting approaches predict ZERO sales, our low-velocity demand forecasting allow us to provide actionable insights into which of your products will sell and when.

Remi Ai’s Demand Forecasting API is an AutoML deep learning API service that utilises best-in-class algorithmic approaches for you supply chain and sales forecasting. Unlike traditional demand planning that relies mostly on consumption and historical sales to build a forecast, Remi Ai encompasses external data streams such as promotional behaviour, weather and global events to more accurately predict future time series. 



Reduction in 

back orders


Increase in

cash flow



Forecasting Methods


Behind our Demand Forecasting APIs are 14 different time series methods, including recurrent neural networks, deep learning, probabilistic, ARIMA and other models to choose from.

But you don’t need to know anything about these models, instead our AutoML service handles all the training, optimisation, testing of the 14 models, so you know you’re guaranteed the best model for your specific data.


This model and its parameters are saved for each of your products, ready to deliver forecasts as new data becomes available.


Different demand forecasting methods


Using best in class probabilistic and deep learning time series models, our APIs hit state of the art of benchmarks, which are up to 50% more accurate than traditional time series methods.  


More accurate 

Some of our clients

Get started with Remi AI