About Astralium: team and research background

Notice: US team with US bachelor's/graduate degrees, 12+ years in the US. In-house engines, golden-sample regression, public self-check cases; data science plus charting rule research. No in-app verdicts; AI does not replace practitioners.

Summary

  • US-based team; members hold US bachelor's and graduate degrees and have studied and worked in the US for more than twelve years.
  • Founders and engineers are the same group; data science and software engineering training plus more than a decade of classical Chinese charting rule research and hands-on practice.
  • In-house charting engines with golden-sample regression; public cases you can check yourself (see delivery standards).
  • No preset lucky/unlucky verdicts in the app; we do not believe AI replaces practitioners.

About Astralium: team and research background.

Notice

Team and location

Astralium is run by a small team based in the United States. Founders and engineers are the same group—charting engines are built and maintained in-house, not outsourced as a black-box service. Members hold US bachelor's and graduate degrees and have each studied and worked in the US for more than twelve years. Training is in data science and software engineering: formal rules, repeatable experiments, automated tests, and written documentation are the default way we ship charting infrastructure.

Charting research

The team has spent more than a decade on classical Chinese charting traditions—Zi Wei, BaZi four pillars, and related systems—through literature comparison, written convention notes, and hands-on casting. The subject is calendar conversion, hour boundaries, rule-set differences, and structural repeatability; calculation conventions for each tradition are documented and maintained alongside engines and packets.

We cast and read charts ourselves and still hold that interpretation requires a practitioner's time, experience, and judgment—large models do not replace that role. AI helps organize literature and follow-up questions; casting and locked conventions belong in infrastructure you can check and regression-test.

Delivery standards

Rigor shows up in checkable, repeatable process—not in-app verdicts or marketing copy:

  • Golden-sample regression—fixed birth/cast inputs against a locked baseline; after any rule or engine change, full recomputation—mismatch blocks release
  • Six-tradition coverage—structural fields and export packets for Zi Wei, BaZi, Western and Vedic astrology, Liu Yao, Qi Men, and related entries under the same verification approach
  • Wheel and packet from one source—the chart view for human eyes; structured packets field-by-field—not chat summaries or OCR
  • Conventions on record—true solar time, day boundaries, and other output-affecting rules written into packets; public cases and self-check steps in our verification article
  • Scope boundary—verification covers chart structure and packets only; lucky/unlucky calls, timing, and useful-god analysis are out of scope—and we do not ship preset readings in the app
  • Operations and billing—paid plans bill through Stripe; see our Terms of Service and Privacy Policy

Why Astralium exists

As generative AI spread, we used it to support literature review and cross-checking rule sets in these traditions. A repeatable finding followed: for the same birth inputs, models drift on palaces, pillars, cycle layers, and other structural fields, and often blend casting and interpretation in one reply. Agent Skills bundle capability into prompts and chat—higher setup cost, hidden charts, poor portability across models (see Skills vs. a charting workflow).

Astralium was scoped to address that: engine casting, structured packets you can verify and copy, and the regression workflow above to lock conventions. Interpretation stays open—to you, a model, or a practitioner (what we build).

FAQ

Where is the Astralium team based, and what is its background?

We operate from the United States. Founders and engineers hold US undergraduate and graduate degrees and have each studied and worked in the US for more than twelve years. Our training is in data science and software engineering, alongside more than a decade of research into classical Chinese charting traditions (Zi Wei, BaZi, and related systems) and long-standing hands-on charting practice.

Does Astralium think AI can replace practitioners?

No. Large models are strong at language and follow-up questions; that is not the same as lineage, experience, or situated judgment. Astralium charts and packages checkable data only—interpretation and readings should stay with practitioners, yourself, or someone you trust.

Why was Astralium built?

After generative AI became widely used, we employed it for literature and rule-set comparison in charting traditions—and repeatedly saw the same issue: for identical birth inputs, models drift on palaces, pillars, cycle layers, and other structural fields, often blending casting and interpretation in one reply. Agent Skills add setup friction and hide the chart. Astralium separates engine casting and structured packets, with regression checks to lock conventions.

How can I judge whether Astralium is trustworthy?

Judge checkable facts, not in-app verdicts: use our public cases—cast the same birth data twice, copy packets, and compare side by side; confirm true solar time, palaces/pillars, and other structural fields against your conventions. Engine rule changes must pass golden-sample regression before release—see our chart verification article. We do not represent any lineage and do not promise lucky/unlucky timing.

What is Astralium?

Astralium charts accurately, then packages structured packets and reading prompts you can paste into ChatGPT or Claude—free to try, no sign-in. Zi Wei, BaZi, Western and Vedic astrology, Liu Yao, and Qi Men—no locked-in verdicts; you and your AI interpret.