4. Lifecycle of a Job

Understand the Job and Choose the Appropriate Type:

This is the initial stage where the job creator determines the nature of the tasks that will comprise the job. They will need to decide on the appropriate type of job based on the data and the type of work required. This could be, for instance, image labeling, text free entry, or a more specialized type like Ask Athena.

Preparing Your Data for Labeling:

After the job type is selected, the next step is preparing the data. This step involves gathering all your data, such as images or texts, and compiling them into a list, often hosted online where they can be easily accessed by the HUMAN network.

Build a Job Manifest:

Once the data is ready, the job creator then builds a job manifest, which is a JSON file that includes detailed instructions for the workers, along with the URLs of the data points from the previous step. The manifest specifies the total number of tasks, the questions for the workers, the minimum and maximum number of times each task should be completed, and more.

Launching a Job:

This step involves instantiating an on-chain job by calling the HMT Escrow Factory contract. The job creator will have to choose the appropriate blockchain network based on their needs. The Escrow Factory generates a new escrow contract for the job, which is then set up by adding the job manifest and funding the contract. Once this is done, the job is successfully launched and is ready to be distributed to the HUMAN network workers.

Monitoring Your Launched Escrows On-Chain:

After the job is launched, the job creator can monitor the progress and status of the escrow contracts. This can be done programmatically using The Graph or HUMAN Protocol SDK, or manually using an on-chain explorer depending on the blockchain network the job was deployed on.

Viewing Results:

The final step is the retrieval and inspection of the results once the job has been completed. Depending on the implementation of the HUMAN Protocol, this may involve pulling data from a specific results endpoint, or interacting with a results dashboard. The results might include accuracy metrics, completed tasks, and potentially location of annotated data.