Hugging Face Email Scraper

Hugging Face Email Scraper - extract model author emails from Hugging Face results

The Hugging Face Email Scraper is an Apify Actor that collects publicly indexed contact emails tied to Hugging Face accounts, model cards, dataset cards and Spaces. You supply keywords, optional location and the email domains you care about, and it returns a clean dataset of ML leads.

Machine-learning practitioners publish addresses in places the Hugging Face Email Scraper is built to reach: the "Citation" and "Contact" sections of a model card, the maintainer block of a dataset card, the README of a Space, and organisation pages for research labs.

This is the tool people mean when they search for a Hugging Face model author email, a way to contact an ML researcher about a checkpoint, or a Hugging Face email extractor that produces a spreadsheet instead of one address at a time.

What the Hugging Face Email Scraper actually reads

The Hugging Face Email Scraper reads only what Google has already indexed on huggingface.co. It builds Google queries with the site: operator, fetches result pages through the Apify GOOGLE_SERP proxy, and extracts emails from titles and snippets.

It does not log into the Hub, call the Hugging Face API, use huggingface_hub, or download a repository. There is no browser, no JavaScript rendering, no authentication and no cookies.

That distinction matters here. A contact address inside a model card is reachable only because Google indexed that card - not because the Actor cloned the repo or read the model's files.

Who the Hugging Face Email Scraper is built for

ML platform vendors selling to model authors, research labs recruiting engineers, compute and inference providers building a launch list, and academic teams looking for collaborators all need the same thing: a repeatable way to turn a research area into contactable authors.

The Hugging Face Email Scraper turns "speech recognition, Gmail addresses" into rows you can filter, dedupe and export as CSV, JSON or Excel, or pull straight from the Apify API.

---

Key features of the Hugging Face Email Scraper

Feature	What it does
Google site: dorking	Queries huggingface.co through the Apify GOOGLE_SERP proxy
Query expansion	Runs base, quoted and intitle: variants plus one variant per query modifier; base queries run first
Domain-filtered extraction	Keeps only emails ending in your customDomains values
Global deduplication	One row per unique email across every query and every page of the run
Obfuscation handling	Understands name [at] domain [dot] com, name (at) domain, name @ domain.com, domain .com, zero-width characters and the full-width ＠
Junk filter	Rejects placeholders such as email@, yourname@, test@, xxx@ and single-character locals
Boundary-correct matching	@gmail.com never matches inside @gmail.company or @gmail.com.br
Soft-wrap repair	Drops a hit that is only the tail of another email in the same result block
Structural parsing	Locates the <h3> title then the smallest surrounding block, instead of relying on Google's CSS class names
Handle resolution	Recovers the huggingface.co/{username} handle from user, organisation and repo URLs
Whole-page fallback	If Google's markup changes, the run degrades to "emails without account details" rather than "no emails"
Block detection	CAPTCHA, "unusual traffic" and consent pages are detected and retried, not counted as empty
Retries and backoff	Up to 3 attempts per page with exponential backoff and a fresh proxy session per request
Requeue of failures	Blocked or failed queries are re-queued once at the end of the run
Async concurrency	An asyncio worker pool with a shared stop signal on maxEmails
Resumable state	Progress is stored in the key-value store keyed by a hash of the input, with throttled saves plus saves on PERSIST_STATE, MIGRATING and ABORTING
Streaming output	Every lead is pushed to the dataset immediately, so partial results survive an abort
Run summary	Logs pages fetched, blocked pages, retries and emails per page

---

How the Hugging Face Email Scraper works

The pipeline behind the Hugging Face Email Scraper is deliberately simple, which is why it runs quickly and is hard to break.

1. The Hugging Face Email Scraper reads your input: keywords, location, email domains and limits.
2. It builds Google queries with the site: operator, for example site:huggingface.co model author "@gmail.com".
3. It fetches Google result pages asynchronously through the Apify GOOGLE_SERP proxy using aiohttp.
4. It parses each result block structurally, finding the <h3> title and the smallest block around it.
5. It extracts emails from the block text with a domain-filtered regex.
6. It deduplicates globally and pushes each lead straight to the Apify dataset.

Why query expansion matters for ML researcher outreach

Google caps a single query at roughly 300 results, so one query has a hard ceiling no matter how many pages you allow.

Query expansion is how the Hugging Face Email Scraper gets past it: each keyword is combined with each email domain in several phrasings, and every variant carries its own result budget.

The default queryModifiers are email, contact, maintainer, author and support - the words that actually appear beside an address in a model card or a dataset card.

What the Hugging Face Email Scraper does not do

It does not log into the Hub, use the Hugging Face API or huggingface_hub, clone repositories, read gated-model access requests, or reach anything that is not already public in Google's index.

It is an independent Apify Actor and is not affiliated with, supported by or endorsed by Hugging Face.

---

Hugging Face Email Scraper input fields

Every field the Hugging Face Email Scraper accepts is optional except keywords. Defaults are exactly those shipped in the Actor's input schema.

Field	Type	Default	Meaning
keywords	array (required)	["machine learning", "model author"]	Search terms describing the Hugging Face accounts you want (research area, task, role)
location	string	""	Optional location phrase added to every query
customDomains	array	["@gmail.com", "@yahoo.com"]	Only emails on these domains are kept; the leading @ is optional
maxEmails	integer 1-10000	20	Stop after this many unique emails
countryCode	string	""	Two-letter country for the search proxy (US, GB, DE...)
expandQueries	boolean	true	Search each keyword x domain pair in several phrasings
queryModifiers	array	["email", "contact", "maintainer", "author", "support"]	Extra words combined with each keyword when expansion is on
maxPagesPerQuery	integer 1-50	30	Page cap per query
maxConcurrency	integer 1-20	5	Parallel queries

Example Hugging Face Email Scraper input

{
  "keywords": ["speech recognition model", "vision language model", "dataset maintainer"],
  "location": "",
  "customDomains": ["@gmail.com", "@ed.ac.uk"],
  "maxEmails": 500,
  "countryCode": "GB",
  "expandQueries": true,
  "queryModifiers": ["email", "contact", "maintainer", "author", "support"],
  "maxPagesPerQuery": 30,
  "maxConcurrency": 5
}

Tuning the Hugging Face Email Scraper for better yield

Task and architecture language works best in the Hugging Face Email Scraper. speech recognition, retrieval augmented generation, quantized llm, instruction tuned and image segmentation dataset match how model and dataset cards are written far better than researcher.

Academic domains are a strong lever here. Adding university domains to customDomains alongside @gmail.com reaches lab authors that consumer-domain filters miss entirely.

Leave expandQueries on. Turning it off gives one query per keyword and domain, which is faster but caps out quickly.

---

Hugging Face Email Scraper output fields

Every dataset item the Hugging Face Email Scraper produces carries all 14 fields below. Fields are never dropped; they are empty or null when Google did not expose the value.

Field	Meaning
network	Platform name
keyword	The keyword that produced the lead
query	The exact Google query used
title	Raw result title
accountName	Account label Google prints (user handle, organisation, or display name)
fullName	Display name parsed from a profile-style title; empty for model or dataset card titles
username	URL-safe Hub handle when one is exposed; otherwise null
profileUrl	Canonical https://huggingface.co/{username} URL when a handle is known; otherwise empty
url	Direct platform link when exposed, else the profile URL
description	Bio or card snippet, cleaned of labels and engagement counters
email	Lower-cased email address
emailDomain	The matched domain, for example @gmail.com
possiblyTruncated	true when Google's snippet ellipsis touched the email - verify before sending
foundAt	ISO 8601 UTC timestamp

Example Hugging Face Email Scraper output

[
  {
    "network": "Hugging Face",
    "keyword": "speech recognition model",
    "query": "site:huggingface.co speech recognition model contact \"@gmail.com\"",
    "title": "ana-vidal/whisper-pt-small · Hugging Face",
    "accountName": "ana-vidal",
    "fullName": "",
    "username": "ana-vidal",
    "profileUrl": "https://huggingface.co/ana-vidal",
    "url": "https://huggingface.co/ana-vidal/whisper-pt-small",
    "description": "Fine-tuned Whisper for European Portuguese. Questions and collaboration: ana.vidal.asr@gmail.com",
    "email": "ana.vidal.asr@gmail.com",
    "emailDomain": "@gmail.com",
    "possiblyTruncated": false,
    "foundAt": "2026-08-31T09:14:22.481Z"
  },
  {
    "network": "Hugging Face",
    "keyword": "dataset maintainer",
    "query": "site:huggingface.co intitle:\"dataset\" maintainer \"@ed.ac.uk\"",
    "title": "northlab/clinical-notes-benchmark · Datasets at Hugging Face",
    "accountName": "northlab",
    "fullName": "",
    "username": "northlab",
    "profileUrl": "https://huggingface.co/northlab",
    "url": "https://huggingface.co/datasets/northlab/clinical-notes-benchmark",
    "description": "Benchmark for clinical note summarisation. Maintainer: j.okafor@ed.ac.uk. Please cite the accompanying ...",
    "email": "j.okafor@ed.ac.uk",
    "emailDomain": "@ed.ac.uk",
    "possiblyTruncated": true,
    "foundAt": "2026-08-31T09:14:29.113Z"
  },
  {
    "network": "Hugging Face",
    "keyword": "vision language model",
    "query": "site:huggingface.co vision language model author email \"@gmail.com\"",
    "title": "Marek Kowalski (marekk) - Hugging Face",
    "accountName": "marekk",
    "fullName": "Marek Kowalski",
    "username": "marekk",
    "profileUrl": "https://huggingface.co/marekk",
    "url": "https://huggingface.co/marekk",
    "description": "Multimodal research. VLM checkpoints and eval harnesses. Reach me at marek.kowalski.vlm@gmail.com",
    "email": "marek.kowalski.vlm@gmail.com",
    "emailDomain": "@gmail.com",
    "possiblyTruncated": false,
    "foundAt": "2026-08-31T09:15:03.902Z"
  }
]

Note how the Hugging Face Email Scraper fills url and profileUrl differently on the first two rows: url points at the model or dataset repository, while profileUrl resolves to the author or organisation account.

---

Use cases for the Hugging Face Email Scraper

Use case	How the Hugging Face Email Scraper helps
ML talent sourcing	Build a list of practitioners by task, modality and framework, with a Hub profile to review first
Research collaboration	Reach the authors of checkpoints and benchmarks adjacent to your own work
ML tooling and infrastructure sales	Contact people already shipping models when launching inference, eval or annotation products
Dataset licensing and provenance	Write to dataset maintainers about terms, corrections or reuse
Academic and lab outreach	Filter to university domains and reach research groups publishing on the Hub
Developer relations	Assemble a launch list of model authors for a beta, grant or credits programme
Ecosystem research	Measure how many authors in a task area publish a public contact channel at all
Conference and workshop programmes	Shortlist speakers and reviewers from a research keyword

Reaching model authors well

Model authors get a lot of noise, so use Hugging Face Email Scraper output carefully. Read the description snippet and open url before writing: naming the actual checkpoint, its base model or its eval results changes the tone of the whole message.

Remember that many addresses on model cards exist for citation and reproducibility questions. A pitch that ignores that context reads as spam even when the address is genuinely public.

Covering the wider ML footprint

Researchers rarely publish in one place. Pair the Hugging Face Email Scraper with the GitHub Email Scraper for training code and with the PyPI Email Scraper for the Python packages that wrap it.

---

Expected results from the Hugging Face Email Scraper

In live test runs against Google, roughly 8 out of 10 parsed Hugging Face results carried an account identity - a Hub handle, an organisation name, or both.

That is one of the stronger ratios in this family, because Hub URLs are structured as huggingface.co/{owner}/{repo} and Google prints the owner in the title. Handle resolution therefore works well on model and dataset cards, not only on profiles.

Hugging Face Email Scraper rows without a handle still carry email, emailDomain, url, description and query. Volume always depends on your keywords, domains and location - no yield is guaranteed.

---

Limitations of the Hugging Face Email Scraper

Read this before planning a campaign around Hugging Face Email Scraper output.

• Publicly indexed emails only. The Hugging Face Email Scraper finds an address only when it is already visible in Google's index. Addresses behind gated-model access forms, private repos or unindexed cards are out of reach.
• No Hub API and no repository access. Model files, config JSON and huggingface_hub responses are never read. Only Google result titles and snippets are parsed.
• Google's ~300-result cap. A single query returns roughly 300 results maximum, which is exactly why expandQueries exists and why turning it off reduces volume.
• possiblyTruncated. When this is true, Google's snippet ellipsis touched the email and it may be cut off. Verify those rows before sending anything.
• username and profileUrl can be empty. They are populated only when the Hub exposes a handle in the result; some Spaces and blog-style pages show only a display name, leaving accountName filled but username null.
• Apify GOOGLE_SERP proxy required. The Actor cannot run without Apify proxy credentials.
• Free plan cap. Free Apify plans are limited to 100 emails per run. Paid plans are uncapped.
• Variable results. Output depends on keywords, domains, location and Google's index, which changes over time. Two runs a month apart will not be identical.

---

Responsible use

Addresses on model and dataset cards are published for research correspondence - citation questions, reproducibility issues, licensing and collaboration. Treat them that way.

If you contact people in the EU or UK, GDPR applies: have a lawful basis, identify yourself, say where you found the address, and honour opt-outs immediately. Respect Hugging Face community norms and any stated contact preference on a card or profile.

Academic addresses deserve extra care. Many are institutional, monitored, and covered by policies that make bulk commercial mail a genuine problem for the recipient.

---

Hugging Face Email Scraper FAQ

Does the Hugging Face Email Scraper use the Hugging Face API?

No. It parses Google search results for pages on huggingface.co that Google has already indexed. There is no API call, no huggingface_hub usage, no login and no repository download.

Does it need a Hugging Face account or token?

No. There is no authentication, no browser, no JavaScript rendering and no cookies. The only credential required is your Apify account's GOOGLE_SERP proxy access.

Where do the emails the Hugging Face Email Scraper returns come from?

From the text Google prints in result titles and snippets - model card contact and citation sections, dataset card maintainer blocks, Space READMEs, and profile or organisation bios.

Can the Hugging Face Email Scraper find authors of gated models?

Only if an address for them is already public in Google's index. It cannot submit access requests or read anything behind a gate.

How many emails can the Hugging Face Email Scraper return?

maxEmails accepts 1 to 10000 and defaults to 20. Free Apify plans are capped at 100 emails per run; paid plans are uncapped. Volume depends on your keywords and domains.

Why is username sometimes null?

Because Google did not expose a Hub handle for that result. The row still carries email, accountName, url and description. On this platform it is the minority case, since owner handles appear in most Hub URLs.

What does possiblyTruncated: true mean?

Google's snippet ellipsis touched the email, so the address may be incomplete. Verify those rows before you use them.

Can I target university domains with the Hugging Face Email Scraper?

Yes, and it works well here. Put them in customDomains, for example ["@ed.ac.uk", "@mit.edu"]. Only emails ending in those domains are kept, and the leading @ is optional.

Does it handle obfuscated addresses like name [at] gmail [dot] com?

Yes. The extractor normalises [at], (at), spaced @, spaced .com, zero-width characters and the full-width ＠, and it filters placeholders such as test@ and yourname@.

Can I run it for a specific country?

Set countryCode to a two-letter code such as US, GB or DE to steer the search proxy, and put a city, region or institution in location to append it to every query.

What happens if a Hugging Face Email Scraper run is interrupted?

Progress is saved in the key-value store keyed by a hash of your input, with saves on PERSIST_STATE, MIGRATING and ABORTING. Leads are pushed to the dataset as they are found, so nothing already collected is lost.

Is the Hugging Face Email Scraper affiliated with Hugging Face?

No. It is an independent Apify Actor, and it is not supported, endorsed or affiliated with Hugging Face.

---

Related Actors

The Hugging Face Email Scraper belongs to a Developer & Technology family of Apify Actors that all work the same way on different platforms. Run several to cover an ecosystem instead of a single site.

Actor	What it collects
Hugging Face Email and Phone Number Scraper	Emails and phone numbers from Hugging Face
Hugging Face Phone Number Scraper	Public phone numbers from Hugging Face
App Store Email Scraper	Public contact emails from App Store
Atlassian Marketplace Email Scraper	Public contact emails from Atlassian Marketplace
Bitbucket Email Scraper	Public contact emails from Bitbucket
Chrome Web Store Email Scraper	Public contact emails from Chrome Web Store
CodePen Email Scraper	Public contact emails from CodePen
Confluence Email Scraper	Public contact emails from Confluence
Dev.to Email Scraper	Public contact emails from DEV Community
Docker Hub Email Scraper	Public contact emails from Docker Hub
Figma Community Email Scraper	Public contact emails from Figma Community
Firefox Add-ons Email Scraper	Public contact emails from Firefox Add-ons
GitHub Email Scraper	Public contact emails from GitHub
GitLab Email Scraper	Public contact emails from GitLab
Google Play Email Scraper	Public contact emails from Google Play
Hashnode Email Scraper	Public contact emails from Hashnode
HubSpot Marketplace Email Scraper	Public contact emails from HubSpot Marketplace
Jira Email Scraper	Public contact emails from Jira
Maven Central Email Scraper	Public contact emails from Maven Central
Microsoft AppSource Email Scraper	Public contact emails from Microsoft AppSource
Salesforce AppExchange Email Scraper	Public contact emails from Salesforce AppExchange
Shopify App Store Email Scraper	Public contact emails from Shopify App Store
Slack App Directory Email Scraper	Public contact emails from Slack App Directory
SourceForge Email Scraper	Public contact emails from SourceForge
Stack Overflow Email Scraper	Public contact emails from Stack Overflow
Unity Asset Store Email Scraper	Public contact emails from Unity Asset Store
Unreal Engine Marketplace Email Scraper	Public contact emails from Unreal Engine Marketplace
WordPress Plugin Directory Email Scraper	Public contact emails from WordPress Plugin Directory
WordPress Theme Directory Email Scraper	Public contact emails from WordPress Theme Directory
Zapier App Directory Email Scraper	Public contact emails from Zapier App Directory
App Store Email and Phone Number Scraper	Emails and phone numbers from App Store
Atlassian Marketplace Email and Phone Number Scraper	Emails and phone numbers from Atlassian Marketplace
Bitbucket Email and Phone Number Scraper	Emails and phone numbers from Bitbucket
Chrome Web Store Email and Phone Number Scraper	Emails and phone numbers from Chrome Web Store
CodePen Email and Phone Number Scraper	Emails and phone numbers from CodePen
Confluence Email and Phone Number Scraper	Emails and phone numbers from Confluence
DEV Community Email and Phone Number Scraper	Emails and phone numbers from DEV Community
Docker Hub Email and Phone Number Scraper	Emails and phone numbers from Docker Hub
Figma Community Email and Phone Number Scraper	Emails and phone numbers from Figma Community
Firefox Add-ons Email and Phone Number Scraper	Emails and phone numbers from Firefox Add-ons
GitHub Email and Phone Number Scraper	Emails and phone numbers from GitHub
GitLab Email and Phone Number Scraper	Emails and phone numbers from GitLab

Leave a review

If the Hugging Face Email Scraper saved you time, please leave a star rating and a short review on the Actor page.

Reviews are how other buyers judge whether a tool works, and they tell us which features to build next.

If something did not work, email neurodata.apify@gmail.com instead - bugs get fixed faster than they get complained about.

Support

Questions, a bug, or a custom build of the Hugging Face Email Scraper for your own domain list? Email neurodata.apify@gmail.com and we will get back to you.