Guided walkthrough

From first project to optimised campaign in AutoRxn.

This walkthrough mirrors the current AutoRxn app. In six steps you’ll go from a blank dashboard to a trained model, suggested next experiments, and a complete project summary you can save and reload.

Launch AutoRxn Back to How it works

What this walkthrough covers

1Set up your project and variables.
2Upload reaction data and generate suggestions.
3Train the model, explore surfaces, and review the summary.

Follow the loop inside the app

Each step below maps directly onto a tab in AutoRxn: Dashboard, Configuration, Reaction Data, Modelling, and Summary.

dashboard

Start on the Dashboard

Create a new project, load an existing one, and sketch the reaction you want to optimise.

Click “Create New Project” to start from scratch, or drag in a saved .json file to continue an existing project.
Use the SketchPad to draw the reaction scheme. This is stored with the project as a visual reference.
When you’re ready, move to the Configuration tab to define your variables.

configuration

Configure Your Project

Define input variables, your optimisation target, and how the model should search the space.

Add each reaction parameter in the Define Input and Target Variables panel.
Choose a type for each variable: continuous input, categorical input, or target variable.
For continuous inputs, set minimum and maximum values over the range you want AutoRxn to explore.
For categorical inputs, list the allowed choices (e.g. ligands, solvents, H₂ sources).
In Configuration Settings, set the project name, acquisition function, and whether to maximise or minimise your target.
Click “Save Configuration” before moving on.

reaction data

Upload Your Reaction Data

Import your existing experiments so AutoRxn can learn a model of your system.

Go to the Reaction Data tab.
Download the template CSV to see the expected column names and format.
Fill in your experimental results so that column names match your variable names from Configuration.
Drag-and-drop the completed CSV into the upload area and click “Load CSV”.
Check the table to confirm that input variables and the target column have all loaded correctly.

suggestions

Generate Suggested Experiments

Ask AutoRxn to propose the next experiments based on your current data and configuration.

Within Reaction Data, switch to the “Suggest New Experiments” tab.
Use the slider to choose how many suggestions you’d like in the next batch.
Click “Generate Experiment Suggestions”.
Review the Suggested Experiments table – each row is a candidate reaction with specific values for ligand, solvent, H₂ source, pressure, temperature, and catalyst loading.
Copy these conditions into your experimental plan or export them as needed.

modelling

Train and Explore the Model

Fit the predictive model and visualise response surfaces, uncertainty, and variable importance.

Go to the Modelling tab and click “Train Predictive Model”.
Once optimisation completes, use the controls at the top to select categorical settings (e.g. ligand, solvent, H₂ source).
Choose two continuous parameters to plot on the axes. AutoRxn will show the predicted response surface for those choices.
Use sliders to set values for any remaining continuous variables.
Inspect the Predicted Response Surface and Prediction Uncertainty plots to see where the model is confident and where more data would be helpful.
Click “Analyze Parameter Importances” to see which variables most strongly influence the target (continuous vs categorical).

summary

Review the Project Summary

Get a single overview of your configuration, data, and suggested next steps.

Open the Summary tab to see top-level project details and notes.
Review the Parameters table to check variable types, ranges, and categorical choices.
Scroll through the Reaction Data table to confirm all experiments are present.
At the bottom, inspect the Suggestions block to see the raw JSON for your current suggested experiments.
When you’re happy, use “Save Project” in the sidebar to export a .json snapshot you can reload later.

Ready to run your own optimisation loop?

Prepare a CSV of your existing experiments, define your variables in Configuration, and follow this walkthrough inside the app. AutoRxn will handle the modelling and suggest where to go next in your reaction space.

Launch app Talk to us