mdai-labs
Open-source installation, solution automation, and field-engineering toolkit
The problem
Customers lacked a consistent way to install a complex open-source platform in on-premises environments. Dependency, versioning, and environment-specific conflicts commonly required one to two days with Jillian as the primary and only field engineer and solutions architect. Learning the MDAI solutions and how to configure them could require another one to two days. Prospects could give up before seeing value, and there was no budget or team capacity for additional hiring.
What I owned
- Architected and built mdai-labs from start to finish.
- Encoded a field-tested solutioning approach so customer enablement did not depend on Jillian's direct availability.
- Open-sourced the repository and provided guidance for community contributions.
How the architecture works.
A field-engineering flywheel encoded as infrastructure
The same reusable solution bundles power customer evaluation, documentation, release validation, engineering feedback, and a later GitOps governance model.
Drag or use arrow keys to pan · pinch or Ctrl/⌘ + scroll to zoom
93%
Read the architecture as text
- Customer + field-engineering problems — encode → Versioned solution bundles + CLI
- Versioned solution bundles + CLI → 5-minute self-service sandbox
- Versioned solution bundles + CLI → Documentation + learning workflows
- Versioned solution bundles + CLI → Benchmark + regression + E2E on every SmartHub build
- Benchmark + regression + E2E on every SmartHub build → Engineering feedback + platform improvements
- Engineering feedback + platform improvements — refine → Versioned solution bundles + CLI
- Versioned solution bundles + CLI — extend → Argo CD centralized governance
Explore the design decisions and capabilities
- Keep the user experience and implementation as simple as practical.
- Install the platform consistently across supported environments.
- Support multi-architecture setup and installation.
- Make workflows completely repeatable through scripting.
- Handle common environment edge cases.
- Support the breadth of platform solutions while enabling deep, advanced examples within a solution area.
- Install the platform through a CLI-like scripted interface.
- Instrument example services or generate representative mock data.
- Accept compatible data streams.
- Provision additional infrastructure required by solution-specific commands.
- Bundle solution configuration and scripts for consistent reuse by the CLI.
- Progress solutions through a reusable field pattern: bare-bones example, static configuration, then advanced automation.
- Demonstrate platform breadth such as centralized ingress and persistent volumes.
- Demonstrate solution depth such as dynamic filtration with tiered groups.
- Run the complete solution suite through automation pipelines for every new SmartHub version.
- Let users evaluate real platform behavior with representative synthetic data before committing to a production-style on-premises installation.
What changed.
- Reduced sandbox evaluation to five minutes through a fully self-service workflow that uses synthetic data to demonstrate the efficacy of MDAI solutions.
- Four of five customer installations completed successfully.
- All 12 internal engineers successfully installed SmartHub and solutions through the CLI.
- Every new SmartHub build triggers the full solution suite for benchmarking, regression testing, and end-to-end testing.
- Let prospects use synthetic data to prove the efficacy of MDAI solutions before encountering production installation complexity.
What it unlocked—and what comes next.
Engineering flywheel
mdai-labs turns field and customer learning into reusable solution bundles, documentation, automated validation, engineering feedback, and inputs for new technical offerings.
- Field and customer problems become repeatable solution bundles.
- Solution bundles become customer learning environments and documentation workflows.
- The same bundles become engineering test fixtures for new platform versions and capabilities.
- Benchmarking, regression, and end-to-end results feed platform improvements.
mdai-labs-compatible Argo CD setup
Extend mdai-labs solution configurations into a GitOps installation and governance model that manages solutions consistently across environments and teams through a centralized Argo CD control plane.
- Manage MDAI through Argo CD.
- Run locally or provision an EKS environment with Terraform.
- Use a GitHub workflow for example AWS deployment.
- Separate development and production applications.
- Keep production in manual-sync safe mode unless automation is explicitly enabled.