關於傳郁國際創智文化

One thesis: Human-Centric AI on the edge.

Ariel Innovations was founded in 2025, in Taiwan, to build Human-Centric AI and Edge AI systems — on the devices people actually have, in the languages they actually speak, with sovereignty and resilience as defaults. Our 2026–2027 roadmap is designed around the institutions our work is meant to serve; we are not yet operating across those markets.

The origin

A company built around a missing piece of the AI stack.

Cloud-first, English-first AI works fine in San Francisco and London. It works much less well in a typhoon shelter, on the floor of a CEATEC pavilion, or in a multilingual training cohort that has to operate across four working languages without losing fidelity.

Ariel Innovations was founded for that gap. The same engineering that earned us Taiwan utility-model patent M678964, Japan utility-model registration 3255721, and a feature on Taiwan Public Television is the technology core we are extending into our disaster-resilience and Civil Training roadmap — the same Human-Centric, multilingual, on-device, sovereign-by-default intelligence, designed for the institutions we believe need it most. We are building toward that audience; we are not yet serving it at scale.

Leadership

The founding team.

Founder & CEO

Sophia Huang 黃舒郁

Sophia Huang, Ed.D. is the Founder and Chief Executive Officer of Ariel Innovations, a Taiwan-founded company building Human-Centric AI, Edge AI, and Security systems. She is co-inventor of Taiwan utility-model patent M678964 (2026–2035) and Japan utility-model registration 3255721 (2026–2035; Japan invention patent pending), which together underpin the company’s Supertitle™ on-device multilingual captioning system — demonstrated on Taiwan’s leading performing-arts stages and featured on Taiwan Public Television (PTS).

Sophia Huang is the founder and CEO of Ariel Innovations and a co-founder of I.P.A.C. Her twenty-year career integrates machine learning, data science, generative AI, human-centric design and research, filmmaking, and storytelling. Before founding Ariel Innovations, Sophia Huang held senior manager roles at Visa, Tesla, LendingClub, and the University of California Office of the President (UCOP).

Education: Ed.D. in Research Methodology, University of San Francisco; Master of Information and Data Science, University of California, Berkeley; MBA, University of San Francisco; MFA in Film, Academy of Art University.

Previously: Visa · Tesla · LendingClub · UC Office of the President

繁體中文 · 黃舒郁 黃舒郁(Sophia Huang)是電影工作者、劇場製作人,也是運用數據敘事的講者—現任傳郁創智科技(Ariel Innovations)執行長、社團法人國際舞臺文化表演藝術交流協會(IPAC)理事長。她將敘事實踐應用於邊緣 AI、表演藝術近用設計與印太區域韌性,主導 Storytelling Supertitles™ 沉浸式智慧型字幕系統(台灣 M678964、日本 3255721 U 新型專利;日本發明專利審查中)。

LinkedIn · /in/huangsophia →

Co-Founder & CTO

Joseph Huang 黃奕勳

Joseph Huang is Co-Founder and Chief Technology Officer of Ariel Innovations. He is co-inventor of Taiwan utility-model patent M678964 (2026–2035) and Japan utility-model registration 3255721 (2026–2035), and architect of the Claude-native engineering stack that underpins the company’s Supertitle™ system and the forthcoming resilience and Civil Training platforms.

Joseph’s career spans senior data and platform-engineering roles across Meta, Uber, Netflix, and Microsoft — building the kind of high-scale systems that the new Edge-AI generation has to inherit and exceed. At Ariel Innovations, his focus is on edge inference, on-device multilingual translation, and the architectural patterns that keep institutional data inside the institution.

Education: Columbia University — dual degrees in Computer Science and Physics.

Selected work: Co-inventor, Taiwan utility-model patent M678964 and Japan utility-model registration 3255721. Published author of the “Artifact + Memory” workflow for teams building with Claude. Lead engineer on the 8,700-line, 19-module Python codebase behind Supertitle™.

Previously: Meta · Uber · Netflix · Microsoft

LinkedIn · /in/joseph-huang-7561392 →

Recognition

What we can publicly point to, today.

Government

iPAS Accredited

Four AI / tech certifications. MOEA / ITRI.

Taiwan Utility-Model Patent

M678964

Registered

Japan Utility-Model Registration

3255721

Registered

Media

Taiwan Public Television

PTS feature on the Supertitle™ stage demonstration.

Platform

Built with Claude

Anthropic infrastructure throughout the engineering stack.

Operating principles

Three principles, in plain language.

1 · Sovereignty by default

If a system can run on the edge, it should. Institutions retain control of their data, their workflows, and their continuity.

2 · Multilingual is not a feature

Every program and platform we are building is authored in four working languages from the first slide. Translation is not delivery; native authoring is.

3 · Lean by design

Claude-native architecture lets us run a lean stack — one production system (Supertitle™) and two on our 2026–2027 roadmap. Operational leanness is a strategic asset, not a limitation.