Hugging Face ML Intern Available as Installable Slack Bot for Workspace ML Tasks

Core Background & Industry Pain Points

ML work often starts in Slack, then jumps to notebooks, Jobs, and Hub. Teams want to evaluate, build datasets, or fine-tune from chat with compute and repos owned by each user or org.

Architecture Highlights & Internals

Hugging Face made open-source ML Intern an installable Slack bot (Chat UI). @ML Intern or a DM runs the same agent as HuggingChat. Each user connects their own HF account for billing and publishing, optionally with MCP. Each thread is a session (up to 200 messages, 10 attachments) with a status card. Compute budgets start at $0 until approved for Jobs/sandboxes; inference bills separately. /btw asks side questions. Self-host via Chat UI Slack docs.

Authoritative Benchmarks & Measured Scores

No official end-to-end success-rate or latency benchmarks. Docs describe capabilities and billing ownership, not scores versus Cursor or Claude Code.

Developer Hands-on Guide

Authorize at huggingface.co/chat/slack/install, add ML Intern, connect HF on first mention. Example: "@ML Intern take the model I linked above and create a Gradio Space for it." Avoid private data in Slack Connect; cancel Jobs on Hub if needed. See official docs.