# Clement Delangue

> 1990– · Entrepreneur, Co-founder of Hugging Face
>
> **Recorded contribution:** Co-founded Hugging Face — open-source ML model hub

## How to use this dossier

Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory.

## 1. Historical orientation

French entrepreneur Clément Delangue co-founded Hugging Face with Julien Chaumond and Thomas Wolf and became CEO. The company began with a conversational app but pivoted toward open-source machine learning, building a hub for models and datasets and supporting libraries such as Transformers. Delangue's contribution is platform, community, and institutional leadership rather than authorship of every hosted model or library. Hugging Face's hub made a model more than a weight file: code revision, configuration, tokenizer, documentation, license, evaluation, and community discussion became a versioned collaborative object. Delangue's institutional role was to align that open ecosystem with hosted services and sustainable company operations.

## 2. The problem inherited

Machine-learning code, weights, datasets, evaluation artifacts, and deployment demos were scattered across papers and private storage, limiting reuse and reproducibility.

## 3. The central contribution

Delangue co-founded and led Hugging Face as it became a major collaboration and distribution platform for open machine-learning artifacts.

## 4. Reconstruct the mechanism

1. Assign versioned repositories and metadata to models, datasets, and applications.
2. Use libraries to normalize loading, tokenization, training, and inference across model families.
3. Store large artifacts with revisions, access controls, licenses, and documentation cards.
4. Let communities discover, test, discuss, and deploy artifacts while moderating unsafe or unlawful use.

## 5. What changed downstream

- Hugging Face lowered barriers to sharing and reusing modern ML models and datasets.
- The hub made openness, documentation, model supply chains, malware scanning, and community governance central infrastructure concerns.
- The platform lowered distribution and discovery costs for open machine learning and created shared conventions for model cards, dataset cards, demos, and reproducible revision identifiers.

## 6. Attribution, limits, and uncertainty

- Hugging Face has three co-founders and many library and community contributors; Delangue is not the creator of Transformers models generally.
- Hosting an artifact does not establish its license, data consent, security, accuracy, or safety; open access can distribute both scrutiny and abuse.
- A popular hub can centralize discovery and trust, hosted artifacts may contain malicious code or unclear licenses, and platform scale does not independently verify a model's data or claims.

## 7. Reconstruction lab

Publish a tiny model repository with pinned code, weights, dataset card, model card, license, checksum, and evaluation. Ask another learner to reproduce it from a clean environment. Audit one model from card to exact revision and executable files, then design quarantine, signing, and takedown rules that preserve legitimate research access.

## 8. Evidence trail

- [Biography of Clément Delangue](https://www.congress.gov/118/meeting/house/116078/witnesses/HHRG-118-SY00-Bio-DelangueC-20230622.pdf) — U.S. Congress
- [Hugging Face documentation](https://huggingface.co/docs) — Hugging Face
- [Hugging Face Hub](https://huggingface.co/) — Hugging Face

---

*Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*