User Guides

Quickstart

Get from sign-in to a completed first job.

Quickstart
Get from sign-in to a completed first job.
Choosing a Service
Match your data and goal to the right workflow.
Data Preparation
Prepare tabular datasets and media manifests for successful uploads.
Jobs and Results
Understand statuses, reports, SHAP, and outputs.
Billing and Credits
How charges, estimates, and refunds work.
Prediction Guide
Use trained jobs for manual and batch prediction.
Troubleshooting
Resolve common user-facing errors quickly.
FAQ
Short answers to the most common platform questions.
Reference
Handbook and Technical Reference

Open the handbook for the full product guide, then use the model-method appendix when you need mathematical details.

Open Handbook Open model-method appendix

Quickstart

Who this is for

Anyone using the platform for the first time and trying to get from raw data to a usable model result without guessing what to do next.

What you will accomplish

  • Choose the correct workflow for your dataset
  • Upload a clean file with the right columns
  • Run a first baseline job with sensible defaults
  • Monitor the job in Workspace
  • Review metrics, explainability, and manuscript outputs
  • Move into prediction or export once the job is complete

The fastest first-job path

1. Start from the homepage or Services

Use the homepage if you already know the kind of problem you want to solve. Use the Services page if you still need to decide between classification, regression, anomaly detection, clustering, survival, or time-series workflows.

2. Match the service to the shape of your data

  • Use cross-sectional workflows when each row is one record
  • Use longitudinal workflows when each subject appears multiple times
  • Use time-series workflows when the order of time matters
  • Use anomaly detection or clustering when you do not have a standard target column

If you are unsure, open the service upload page and compare the required columns shown there before uploading.

3. Prepare one clean input file

Use a supported tabular dataset file, such as CSV/TSV, Excel, Parquet, Feather, JSON/JSONL, Stata, SPSS, SAS, or R data. Do not upload .pkl or .pickle files; pickle deserialization is unsafe for untrusted uploads and is rejected by the app. Before upload, confirm: - column names are readable and unique - the target column exists when the workflow requires one - date and time fields are consistently formatted - identifier columns are included only when the workflow needs them

4. Preview readiness before training configuration

Map the required roles, choose feature columns, and set the positive class for classification workflows before clicking Preview Data Readiness. Training configuration, model selection, tuning, cost confirmation, and launch controls appear after readiness succeeds.

5. Build the first comparison set

Start with the balanced comparison strategy if you want a practical first run. The model catalog remains visible after readiness so you can broaden, narrow, or target model families before launch. Leave advanced tuning controls closed unless your study protocol calls for manual overrides.

6. Review cost before you launch

Before submitting the job: - confirm the selected service - confirm the target column - read the estimated charge - confirm whether SHAP or manuscript generation will be triggered later as separate add-ons

7. Submit the training job

Once submitted, the job moves into Workspace where you can monitor: - queue state - live status updates - latest logs - produced artifacts - follow-on actions like SHAP, manuscript generation, or prediction

8. Read the results in the right place

  • Use the job detail page for one-job review, billing context, SHAP, and generation actions
  • Use the AI Report tab on job detail for embedded manuscript review, provenance, figures, tables, export, and version checks
  • Use the focused reader route only when you want the report on its own page
  • Use prediction screens only after the job has produced a usable trained artifact

What a successful first run should give you

  • a best-model summary
  • model comparison metrics appropriate for the selected service
  • explainability artifacts when supported
  • a manuscript and executive brief when generated
  • prediction entry points for supported services

Common first-run mistakes

  • choosing a service that does not match the data shape
  • using the wrong target column
  • uploading identifier columns as predictive features
  • changing tuning controls before you have a baseline
  • assuming SHAP or manuscript generation is part of every completed training run

Recommended next step after your first baseline

After the first run succeeds: 1. Review metrics and compare models 2. Launch SHAP from the job page if you need explainability 3. Generate the manuscript outputs 4. Tighten preprocessing, tuning, or model selection after reviewing the baseline

Related guides

  • Choosing a Service
  • Data Preparation
  • Jobs and Results
  • Billing and Credits