Context
Lobe was incubated in the Office of the CTO to make machine learning approachable through a free desktop app. People could train a custom model by providing examples, then use that model in their own application.
Role & Scope
I guided the front-end team for Lobe and led migration of the Lobe web application and services from cloud infrastructure to an Electron desktop architecture.
I worked with engineering and product partners through migration, interactive UX development, and release readiness as Lobe moved from incubation to broader Microsoft ownership.
Constraints
- The desktop product needed to hide the complexity of model training without reducing control and clarity for users.
- Migrating from cloud-hosted architecture to desktop had to preserve core workflows for training, evaluation, and model management.
- Real-time interfaces had to remain responsive while users iterated on examples and model behavior.
Key Decisions & Contributions
- I led the shift to an Electron-based desktop architecture to support shipping Lobe as a free offering.
- I drove highly interactive UX for model training, evaluation, and management so advanced machine learning steps stayed understandable in a visual workflow.
- I aligned front-end direction with release goals needed for public launch and organization transition.
Outcomes
- Lobe shipped as a free desktop application.
- The product reached public release.
- Lobe graduated from incubation into the Microsoft Power Platform organization.
Reflection
This project validated a recurring pattern in technical product design: difficult systems become usable when feedback is immediate and decisions are legible. My focus remains building interfaces that make complex capabilities practical for everyday teams.
Professional contact
If this project work is relevant to your team, connect with me on LinkedIn.
LinkedIn