TurboML

TurboML brings streaming data, feature engineering, model training, and monitoring into one platform. I joined when backend development had started but there was no app, and designed the experience from the first flows through launch.

My scope covered the product UI, design system, brand identity, website, and launch materials.

A dataset could pass through several tools before becoming part of a model. Different controls and conventions left engineers and data scientists piecing together what had happened along the way. Streaming data added another question: what is the system doing right now?

The backend was changing as I designed. I worked closely with the founders, developers, and internal ML team to give the workflow a clear structure while preserving the technical controls people needed.

The team shipped the first MVP in four months.

I organized the app around four connected stages: ingest data, create features, build models, and monitor performance. Source setup made it possible to configure a connection, verify incoming data, and manage datasets in sequence. Feature engineering offered both code and visual paths, with previews to inspect the results.

Live updates and logs exposed system state. A pipeline view connected sources, feature stores, and models in one diagram, helping users follow how the pieces worked together. A shared component system kept the growing platform consistent.