Maxis Workspaces - Deploy
My role
UX/UI designer
and researcher
Team
UX/UI lead
Product manager
Architects
Status
Shipped Q1 2023
PROJECT OVERVIEW
What is Workspaces?
Maxis Workspaces is an internal tool within the Maxis Product Suite at General Motors, serving over 3,000 unique users across different roles. It is designed to help various user types integrate and leverage data in dynamic workspaces. Data engineers use it to gather raw data from multiple sources and create one-off datasets, while data scientists and engineers enhance and refine these datasets through the development of 360s data models. Analysts and data scientists work with raw data sourced from both internal and external datasets to support analysis in the GM data lake. Users create sandboxes for collaboration, work with Jupyter notebooks, access popular datasets, and design workflows that streamline their daily tasks.
About low code app deployments
The Low Code Apps feature in Maxis Workspaces simplifies the traditionally complex deployment process for Dash applications by reducing a lengthy, nine-step workflow down to three main steps. It targets two personas: developers and users. Developers build their apps in Jupyter and then use an integrated widget to package their files, automatically creating the necessary configurations, like main.py
and requirements.txt
, which are stored securely.
The streamlined deployment process eliminates the need for developers to understand complex backend operations or go through extensive approval steps. Once packaged, a deployment wizard guides them through three quick steps: selecting the package, applying security credentials, and choosing a publication location within Maxis Consumers. For MVP, there will be default resource settings, ensuring deployments are adequately supported without the need for advanced technical knowledge.
01/
Why are the users struggling?
the problem
traditional dash app deployment is overly complicated
Today in GM, deploying a dash app requires navigating a 40-page document that outlines a complex process. These steps involve multiple stages of setup. Each stage requires significant preparation, approvals, and a steep learning curve for various internal processes, making the deployment process time-consuming and inefficient, which detracts from the core work of developing the app, creating frustration and slowing productivity.
02/
Understanding past research/technical
Before moving further with the project, it was crucial for us to analyze the previous research conducted and undestand the backend of provisioning a low code application. The following is the information we gathered:
user research
Pain Points
• Reluctancy to transition to Automate from their previous tool Oozie
• Users question the reliability of Automating Deployments vs Oozie
Feature Requests
• Enhanced edit capabilities for Workflows
• Recieve user friendly error messages for failed workflows
• Recieve email notifiications re failed jobs
deployment user/code flow
The technical workflow is original and innovative, reflected in the comprehensive diagrams illustrated by our development team. This would later help us understand what needed to be a part of the UI, the kinds of metadata and the larger flow that the user’s code would be going through. These whiteboarding sessions helped us solidify a general idea of the feature.
emerging opportunities
streamlined three-step workflow
Automated Environment Setup and Security Configuration: The Low Code Apps flow automates backend setup tasks, allowing developers to focus solely on application creation without needing to understand complex infrastructure. This automation provides a faster, more efficient deployment process.
Simplified Deployment Interface with Preset Configurations: Our user flow uses a Jupyter widget to package the application files automatically and guide developers through a simple form-based wizard. This low-code interface reduces technical barriers, allowing non-expert users to deploy their apps confidently with preset configurations, reducing the need for extensive technical training.
Integrated Publishing and Monitoring within Maxis: Eliminate separate approval and review steps by directly integrating the deployment into Maxis Consumers, where apps are automatically published and accessible. This integrated approach provides users with real-time deployment status and immediate access to their apps, enhancing transparency and reducing the delay associated with multi-step approvals.
03/
User experience flow
analysis & mapping
After our understanding of the feature requirements, we created a user flow using Miro to use as our reference before designing. Creating these user maps individually and then coming together to share protects us from each other’s bias of how the flow should be constructed. After sharing our individual flows, we agreed on the main steps and created a final version.
04/
Getting to the final product.
final designs
With the confirmation of the user flow, our the design team began looking for UI inspiration before creating low-fidelity mockups.
The UX of the feature would require a seamless and reliable way for our users to be able to deploy a low code app. This meant that the user needs to go into their workspace, fill out a form that automates a deployment and creates a link to a live application that will be hosted on Maxis Consumers. Once deployed, we needed to allow our users to monitor, edit, and delete their deployments or errors. Visually, this would be visible in a tab where they can monitor and manage their deployment, but also create a new deployment.
Screen: Deploy Landing Tab
Provide a tab within the Workspace so users can deploy low code applications.
Screen: Create Application Name & Description
Screen: Deployment Form
Users can complete an easy to use form to create low code deployments. In a three-step form users will be asked to assign admins, secure & configure, before confirming their aplication deployment.
06/
Evaluating success through metrics
outcomes
The deploy feature was launched at the end Q1 in 2023
Next Project
Data Catalog