Setting up a Project
This guide will show Data Science Leaders how to set up a project in Vectice web UI.
Setting up a project is the first step to empowering your business and teams to accelerate the impact of AI within your organization. Projects allow data science leaders to manage and monitor their team's workflows and gain visibility into data science initiatives. Furthermore, data scientists can continue working with familiar tools while capturing milestones and reporting key findings to colleagues and stakeholders.
In this guide, we will walk you through setting up a project.
Create Project
To create a project:
Select your desired Workspace.
On the Projects page, select + Create Project.
Enter your project’s name and description.
Determine whether you want to
Create a Blank Project: As the name suggests, you will start with an empty Project.
Create from a Default Template: A project that follows the CRISP-DM methodology.
Create a Sample Tutorial Project: A project that enables you to learn Vectice.
Lastly, select Create.
Once you create a project, you will automatically become the Project owner. You can reassign the project ownership to any Workspace member.
🎉 You have created a project! Now let’s move on to Phases.
Create Phases
Phases allow you to break down the project into milestones or objectives to organize the project, enforce best practices, allow consistency, and capture knowledge. You can document the goals, the assets, and the outcomes of the work, along with the status. The phases reflect the real-life phases of the project lifecycle, for example, Business Understanding, Data Preparation, Modeling, Evaluation, and Deployment.
To create a phase, you’ll need to:
Select the Phases tab.
Select + Create Phase or click the Create button on the left sidebar next to Phases.
Enter the Phase name, then select Create.
🎉 Great! You have now created a Phase for your Project. Next, we will define the steps needed to execute the Phase.
Documentation
Documentation allows you to describe phase requirements and summarize the outcome of the phase by constructively evaluating accomplishments and results that need improvement. For example, in terms of the model, you can highlight how a model has improved by 35%, but it may still contain a percentage of bias due to data integrity issues.
You can add the phase requirements and summary by clicking on the Documentation tab.
For a detailed walkthrough on how to utilize the documentation tool, view the Summarizing Outcomes Guide.
Iterations
Data Science is an iterative process that arrives at a refined outcome (like a Machine Learning model) over time. Iterations can refer to updating and improving the model's parameters after training, testing, and validating. Those steady and continual improvements are needed to build the best models.
The Iterations tab will present all iterative development work on the Phase. You can track and manage iterations once they appear.
You must define at least one Step so that Data Scientists can begin syncing their iterative development for the Phase steps. They will be able to share their updates and metadata with each step while developing.
Use the Vectice API to get started on iterative development. For a walkthrough on how to sync your iterative development to Vectice using the API, visit the Iterative Development Guide.
You can also read the API docs to learn more about the Vectice API.
Create Steps
Defining the sequence of steps and expected outcomes needed to complete the project phase will guide your team toward success.
To create steps for your project’s Phase, complete the following:
Select the Steps tab on the Phases page.
Select Create Step to begin defining the steps needed to complete the Phase.
Enter the Step Name and Description, then select Create Step.
You have created a step in the Phase! To add more steps, complete these instructions again.
Set your team up for success by ensuring each step is descriptive and includes success metrics or expected outcomes.
Reviews
Reviews allow you to receive feedback from stakeholders and subject-matter experts on a Phase. Read the reviews to stay up to date on the conversation and feedback.
To request a review, you'll need to:
Select the Reviews tab.
Click the “Request a Review” button.
Select your Reviewer. Leaving a message for the reviewer is optional.
Click “Send Request”.
The reviewer will receive a notification of your request. They will have the option to leave comments, request changes, and approve the review.
To request a review, there must be a phase owner, and that owner is only authorized to make a request for review.
Datasets
Datasets reflect the datasets used for analysis, cleaning, model training, and validation. Datasets from various source systems can be registered in the catalog along with their metadata. Vectice will automatically version the metadata of datasets to enable lineage.
You can view, search, and filter all datasets (and their various versions) used in the Project along with their type and other information by clicking Datasets.
Models
Models represent the machine learning models created and trained during the modeling process. The model's metadata, hyper-parameters, and the metrics from the training or production process are stored. Models are versioned and tagged with their deployment environment within Vectice.
You can view, search, and filter all models (and their various versions) used in the Project along with their type and other information by clicking Models.
Activity
The activity page provides an overview of what's happening on the project. You can filter by Contributors or Activity type.
Project Settings
| Settings | Description |
|---|---|
Move Project | Moving a project to another workspace moves all phases, models, and associated datasets. You can only move projects to workspaces where you are a member. |
Duplicate Project | Duplicating a Project creates a new Project while preserving all the Phases, Documentation, and Steps in the new Project. Iterations, models, and datasets are not copied to the new Project. |
Project owners and Workspace Admins can transfer project ownership to another Workspace member. Only Workspace Admins can undo this action. | |
Deleting a project also deletes all phases, models, and datasets associated with it. This action is irreversible. |
Last updated
