Hugging Face Launches 'ML Intern,' a Chat-Based Agent That Runs ML Experiments Autonomously

Hugging Face has launched ML Intern, an AI assistant embedded in its chatbot interface that lets users run machine learning experiments without writing code or having ML expertise. The tool is designed to turn a plain-language description of an idea into a fully executed experiment, from data collection to a working demo.

According to Hugging Face, the workflow starts with a conversation: a user describes what they want to build, and ML Intern searches the Hugging Face Hub, GitHub, and the general web to identify relevant models, datasets, and tools. Before starting any compute-intensive work, the assistant estimates the cost of the experiment and proposes a budget. Once a user approves that budget, ML Intern is constrained not to exceed it.

After approval, the system operates independently. Hugging Face says it can generate datasets, train models, monitor active training jobs, upload results back to the Hub, write summary reports, and build interactive demos. Each training run is tracked through its own dedicated dashboard, giving users visibility into progress without requiring them to manage the underlying infrastructure.

Hugging Face cites one example from a demo video in which a training run lasted roughly six hours and cost less than $0.50 — though this figure comes from the company's own demonstration rather than independent verification, and actual costs will vary by model size, dataset, and compute requirements.

The launch effectively lowers the technical entry point for building and training models on the Hugging Face platform, extending automation that previously required ML engineering knowledge to a broader base of non-technical users.

The release comes as Hugging Face is reportedly in the process of being acquired by Nvidia, a deal the source article references but does not detail with specific terms. Nvidia CEO Jensen Huang has stated, according to the report, that Hugging Face will remain open and hardware-neutral following the transition. Neither company has published detailed terms of the acquisition alongside this launch, and pricing or availability details for ML Intern beyond the demo example have not been disclosed.

What this means

ML Intern fits a broader industry pattern of wrapping complex technical workflows — here, end-to-end ML experimentation — inside conversational agents that handle search, execution, and reporting autonomously. The budget-capping mechanism is notable: it addresses a real pain point for non-experts, who often have no intuition for how much a training run will cost before it's too late to stop it.

The more consequential story may be the Nvidia acquisition context. If Hugging Face is indeed being absorbed by Nvidia, a tool like ML Intern that automatically selects models and compute resources raises questions about neutrality — whether such an assistant would preferentially route jobs toward Nvidia hardware or Nvidia-hosted infrastructure. Huang's reported commitment to keeping the platform open is worth watching closely as concrete product decisions, like this one, roll out under new ownership. Until Hugging Face publishes independent benchmarks and cost data beyond a single demo example, claims about affordability and capability should be treated as preliminary.