Source-linked jobs, companies and research signals — in one explorable map.
Public website · 中文介绍 · How it works · Contribute a source
Interface concept illustration, not a live screenshot. An IPPLab project.
Finding a path into embodied AI often means jumping between career pages, company websites, release notes and social platforms. Embodied Compass brings those public signals together, while keeping the original source and collection date visible.
Start with a role. Explore its skills. Follow the company. Read the release. Then return to the original source before you act.
| Explore | What you can do | What the data means |
|---|---|---|
| Opportunities | Search by role, company or skill; filter by city, specialty, role type and official source | Parsed public listings, not guaranteed openings |
| Companies | Browse the ecosystem and open company or career pages | A source directory, not proof of current hiring |
| Research | Follow source-linked open-source releases and research or industry leads | A release is not independent performance validation |
| Social signals | Read separately labelled search-index leads when available | Unverified leads; excluded from job counts |
The public interface supports Chinese and English. Source text stays in its original language. No sign-in is needed. Personal match scores and private saved links are excluded. The current collector focuses on China, with Shenzhen-first career sources and some adjacent AI roles; this is not a global market census.
Python 3 is enough to serve the included public snapshot. No frontend build or package installation is needed for the public website.
git clone https://github.com/codeofwhite/embodied-intel-radar.git
cd embodied-intel-radar
python3 -m http.server 8080 --directory site --bind 127.0.0.1Open http://127.0.0.1:8080. This reads the committed snapshot; it does not start a collector or connect to the author's private database.
Public career pages / company sources / release and research indexes
↓
Existing Python collectors (robots-aware, rate-limited)
↓
Field-allowlisted public JSON export
↓
GitHub Pages · search, filter, follow the source
GitHub Actions attempts a refresh every day at 08:00 China time (00:00 UTC). The page shows actual collection dates; schedules can be delayed or disabled by GitHub. Failed required collection or invalid exports leave the last published snapshot intact. Optional sections retain their own timestamps if collection fails. No paid model API is required.
For your own deployment, choose GitHub Actions in Settings → Pages, then
run Publish Compass. The first deployment requires this repository setting.
See the deployment guide for permissions, failure behavior,
and refresh commands. The workflow checks for the hosted runner's existing
requests dependency and fails explicitly if it is unavailable.
- A page that loads is not necessarily an open job. Check deadlines and the employer's original listing. Missing dates, old posts and extraction conflicts remain visible.
- Counts describe collected samples. Search coverage and parsers are limited. Social leads and company directory entries are not counted as jobs.
- Source tiers describe provenance. They do not certify accuracy. Automated extraction and heuristic analysis can be wrong.
- A snapshot is not a trend. The public refresh workflow does not accumulate a multi-day database or claim hiring growth, recruiting probability or investment value.
- Public means public sources. No login, CAPTCHA or access-control bypassing. Links and short extracted descriptions remain attributed to their sources.
Methodology · Reading guide · Collector details
The standalone public UI lives in site/. The retained React/Worker/D1 application
in app/ is a separate private deployment with inbox and refresh features. Its
routes do not provide multi-user isolation; do not expose the full private service
anonymously. The public website never calls those APIs.
With the existing Python 3 + requests development environment:
python3 -m unittest discover -s collector/tests -v
node --check site/app.jsTo refresh locally, follow the exact public collection commands.
Collection writes local data and makes external requests. radar_refresh.py also
includes optional notification/private publishing steps, so the public workflow
calls only the specific collectors it needs.
Good contributions are concrete: a public company career URL, a broken source, a reproducible parser bug, better source labels, or an accessibility improvement.
Open an issue with the original URL, what you expected, and what you observed. Please omit resumes, private messages, contact details and access credentials.
Created as an IPPLab project. Project branding is independent of the lab name; IPPLab attribution appears on the website and project illustration.
Third-party code retains its notices: third-party notices. A repository-wide license has not yet been selected. Public visibility does not by itself grant unrestricted redistribution rights. Third-party source content remains subject to its original terms.