Software / AI Platform
Vertex
Vertex
About
Full-stack development and interface design for an AI model training platform. Vertex enables ML teams to manage datasets, run experiments, and deploy models through an intuitive visual interface.
Client
Vertex AI Labs
Date
August — December 2023
Services
Software / AI Platform
Project Gallery
Vertex pipeline builder
Visual pipeline builder interface
Vertex training dashboard
Real-time training metrics dashboard
Vertex experiment comparison
Experiment comparison view
Our Approach
- 01Mapped the entire model lifecycle from data ingestion to deployment, identifying 12 key interaction points that needed streamlining.
- 02Built a drag-and-drop pipeline builder that abstracts away infrastructure complexity while giving experts full access to configuration.
- 03Implemented real-time training visualisation with WebGL-powered charts that update without jank, even at 100+ metrics per second.
- 04Designed a collaboration layer with version control for experiments, allowing teams to fork, compare, and merge training runs.
Challenges
- ✕The pipeline builder had to handle recursive branching logic — a significant engineering challenge that required rewriting the graph engine twice.
- ✕Performance at scale: some users train models across 100+ GPUs, generating terabytes of log data. Keeping the UI responsive was a constant battle.
- ✕Balancing simplicity for newcomers with power features for ML engineers — we nearly shipped an overly simplified v1 before user testing caught it.
Client Feedback
“Vertex has changed how our team operates. What used to take three days of scripting now takes twenty minutes in the visual interface.”