Use your own data to train your own model


Include your own forecasting models in your app in a couple of minutes.

Let Nixtla's API do all the heavy lifting. No servers to manage, no configuration, no headaches. It just works.

1. 🔑 Get your Token

Click here to authenticate yourself and get a token. You should see the following screen.

2. 👩‍💻 Upload your data

Upload your time series data to a cloud provider and get the url associated. Your data must have the following format. The unique_id column identifies each time series, ds identifies the date (or timestamp) of each observation, and finally y contains the target variable you want to forecast for each time series.

3. 🤖 Choose your model

We have a large collection of models you can use to fit on your data.

🧠 Deep Learning


  • nhits(Nixtla)
  • nbeats (ElementAI)
  • nbeatsx(Nixtla)
  • tft (Google)
  • deep_ar (AWS)
  • lstm(Schmidhuber)
  • esrnn(Smyl)
  • DeepAR(Amazon)


  • Transformer Models (Google)
  • DeepVar (Comming Soon)

Statistical Models

  • arima: AutoARIMA Model
  • seasonal_exponential_smoothing
  • prophet(Facebook)
  • complex_es
  • ets (Top 5 in M3)`

4. 🚀 Train your model in your fav language.

Make a REST call. In the following example, we use Node, but you can use the programming language of your choice.

5. 🔮 Forecasts with your trained model

Once the model is fitted, you can start to produce forecasts using a REST call in two simple steps. Just follow the following instructions.