Hugging Face

Searches Hugging Face for datasets, filters rows by criteria, and exports a parquet or CSV slice. Handles metadata, pagination, and stats lookups without local setup.

Sets up and runs supervised fine-tuning or preference-based training (DPO, GRPO) on Hugging Face infrastructure, then converts the result to GGUF for local use or deployment.

Trains object detection, classification, or segmentation models on your dataset using Hugging Face. Handles data prep, model selection, and outputs a ready-to-deploy checkpoint.

Writes Gradio code that runs on Hugging Face's free ZeroGPU tier, handling the quirks of process isolation and pickle serialization so your demo runs when you hit deploy.

Names the top open-source models ranked by benchmark scores for your specific task—text generation, image, coding, classification—with their speed and memory tradeoffs side by side.

Takes a research paper, formats it for Hugging Face Hub, links it to your models and datasets, and publishes a professional page with proper authorship and citations.

Use to select models to run locally with llama.cpp and GGUF on CPU, Mac Metal, CUDA, or ROCm. Covers finding GGUFs, quant selection, running servers, exact GGUF file lookup, conversion, and OpenAI-compatible local serving.

Wraps the Hugging Face CLI to authenticate, download models and datasets, upload your own work, and manage repos—without leaving Claude.

Runs standardized evals against Hugging Face models using inspect-ai and lighteval, comparing performance across backends (vLLM, Transformers, accelerate) without cloud costs.

Scaffolds a Gradio app with components, event handlers, and layout that runs immediately. Handle text, images, or chat without writing boilerplate.

Generates a reusable script that fetches models, datasets, or inference results from Hugging Face and chains them into your workflow — no API boilerplate to memorize.

Takes your labeled examples and trains a sentence-transformer model to embed text the way your domain needs it — for retrieval, similarity, clustering, or ranking tasks.

Fetches the full paper from Hugging Face or arXiv, extracts authors and linked code repos, datasets, and model weights, then summarizes the key contribution in plain English.

Connects to your training loop via Python API, logs metrics as they run, fires alerts when things drift, and syncs results to a live dashboard for quick spot-checks without stopping the job.

Wraps Transformers.js to let you classify text, detect objects in images, transcribe audio, or translate—all in JavaScript, no API calls or setup infrastructure needed.

Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.