Athina AI enables you to launch fully managed annotation projects to collect high-quality labeled data. This guide walks through the full setup process—from project creation to annotator assignment.

​
Step 1: Create a Project

Navigate to the Annotations page and click the + Create project button at the top right. Create Annotation Project On the creation page, fill in the required project information:
  • Name – Internal name of your annotation project.
  • Description – Optional. Helps team members understand the scope.
  • Instructions – Optional guidelines shown to annotators for consistent labeling.
Project Creation Form

​
Step 2: Select Dataset

Choose a dataset from the dropdown.
Currently, only existing datasets can be selected.
  • Supported file formats for datasets: .jsonl, .json, and .csv
  • If needed, upload the dataset from the Datasets page beforehand.

​
Step 3: Define Annotation View

Annotation View determines what annotators see and how they interact with the data.
  • You can select a template (predefined configuration) or create a new view.
  • Define:
    • Which fields are visible (e.g., input, output, metadata).
    • Whether fields are editable, required, or markdown-enabled.
    • Labels to be collected (e.g., category, number, comment).
View configurations can be reused across multiple projects.

​
Step 4: Set Completion Criteria

Define how many annotators must label each datapoint.
  • Minimum required annotations per datapoint:
    Determines how many independent annotations are needed before a datapoint is considered complete.
  • Toggle: Skip datapoint if minimum requirement is met – allows annotators to skip fully-labeled entries.
If you set this value to 2, each datapoint must be labeled by 2 different annotators to be marked as complete.

​
Step 5: Choose Assignment Method

Select how to distribute annotation workload:
  • Entire dataset – All annotators label the full dataset.
  • Split into subsets – Automatically or manually divide the dataset and assign different subsets to annotators.
Use subsets if you want to avoid duplicate annotation or manage workloads among many contributors.

​
Step 6: Assign Annotators

  • Search and add your team members as annotators.
  • Depending on the assignment method, they will receive:
    • All datapoints (entire dataset) or
    • Only their subset (split mode)

​
Step 7: Launch

When you’re done configuring, click Launch project. Annotators will be notified and can begin labeling immediately.