About Soma Furuie — English to Japanese Patent Translator

Freelance patent translator. English to Japanese. Based in Kyoto.

My turnaround (1.5–3 business days, depending on document size — see Services for the breakdown) isn’t about speed. It’s a workflow built to deliver filing-ready Japanese with margin before your deadline.

Background

Joined a Japanese patent translation company in 2019. Spent 5 years and 9 months in-house, at 2 million words of patent content per year. 80% of that time was reviewing translations by colleagues and external contractors; the rest I translated myself.

With ongoing freelance work since, I have now passed 13 million words of patent content overall: 10 million reviewed and 3 million translated.

That reviewer experience is the backbone of the translation work I do today. I learned, at scale, where patent translations go wrong:

  • Terminology drift
  • Omitted reference signs
  • Mis-rendered numerical values
  • Inverted modifier scopes
  • Claim-figure mismatches

In 2025, I transitioned to full-time freelance patent translation. Translation is now my primary deliverable — the reviewer’s eye runs underneath every sentence I produce.

Built on Language, Not Engineering

I came from Sophia University’s Faculty of Foreign Studies, English Department. A language background, not engineering. That focus is what makes my translation work strong.

Engineers-turned-translators can read circuits and machinery. But the hardest parts of patent translation are grammatical, structural, and interpretive. A patent claim is a legal sentence first and a technical description second.

Reading source English carefully and rendering it precisely in Japanese is what I trained for. Technical fluency followed through years of immersion in patent documents.

My Workflow

I’ve developed my own translation system, and continue to extend it with every project. 26,000+ lines of Python, designed specifically for patent work. Claude AI on AWS Bedrock for the first draft. I check every sentence of the translation by hand.

What’s behind it:

  • A glossary system that locks client-specific terms and prevents bleed across projects
  • Project-start term verification so terminology preferences are agreed upfront
  • Layered automated and AI-assisted quality checks before manual review
  • Translation memory that compounds consistency across your filings
  • A translation-notes document I prepare alongside the translation, listing source-text issues — typos in the original, reference-sign mismatches, and passages I had to interpret

What this changes about my work: the drafting layer — typing, terminology enforcement, grammatical normalization — happens before I open the file. The remaining work is the work AI can’t do: interpreting source intent, preserving claim scope in translation, catching antecedent (pronoun-reference) traps the system can flag but not resolve.

AI doesn’t replace patent translation expertise. It concentrates my time on the judgment a translation depends on.

No AI training on your content. No data retention by the AI model. Processed on AWS Bedrock’s Tokyo region (ap-northeast-1). NDA available on request.

How it all works in detail →

Why Patents

A mistranslated term can shift a patent’s scope. The work demands meticulous, terminology-sensitive thinking. My clients are patent attorneys, IP departments, and translation agencies who cannot afford ambiguity. That responsibility is the work I want.

Working Style

I work best with clients who view translation as part of the patent filing process. Clients who value early identification of source issues, technical accuracy, terminology consistency, and a record of what I caught and how I read the ambiguous parts.

Not a fit: rate-driven volume contracts, projects requiring specific CAT (computer-assisted translation) tool mandates, work outside patent documentation.

Professional Profiles

Get in Touch

Inquiries from patent attorneys, IP departments, and translation agencies welcome.

Contact →