AutoML Experiment Tracker

In Development

A collaborative platform for ML teams to track experiments, compare model performance, and share insights through interactive visualizations.

Source Code

Problem

Existing tools were either too complex (MLflow) or too simple (spreadsheets). Needed something that balanced power with usability.

Architecture

Frontend uses React with a custom canvas-based visualization engine. Backend is FastAPI with PostgreSQL and async job processing via Celery.

Challenges

Rendering thousands of data points in real-time without blocking the main thread. Solved using OffscreenCanvas and a custom virtual scrolling implementation.

Lessons Learned

Performance is a feature. Users will forgive missing functionality, but they won't forgive a laggy interface when they're iterating quickly.

Tech Stack

TypeScriptReactFastAPIPostgreSQLRedisDocker

Details

Timeline2023

StatusIn Development