ML-Data-Analyst

User facing UI. Main agent orchestrating sub-agents to deep dive into ML datasets.

Inspiration

This project will integrate into my own start-up: https://rapidreview.io

What it does

Finds you the best ML dataset on Huggingface for you ML tasks

How we built it

Using Codex, and Claude Code

Challenges we ran into

Classic bugs in agentic development

Accomplishments that we're proud of

Scaling and User Experience. Each dataset that we do a deep dive in gets a dedicated agent for better UX and a dedicated click house worker to ensure separation.

What we learned

It is not that hard to inspect real datasets with ClickHouse.

What's next for ML-Data-Analyst

Integration into autonomous ML training workflows. See https://rapidreview.io for more info. One downstream automation process is hybrid dataset generation, merging existing datasets, as well as expanding coverage beyond HuggingFace.