Written by
A freelancer’s work is a series of overlapping, often long-running collaborations, so periods below are not strict employment blocks. Rather, they emphasise points when a particular kind of work became important.
2004–2008 · Technical foundations
I started with manuals, web content, press releases, and promotional material for a PC hardware manufacturer.
From 2008, work through SDL expanded that into large software and support projects, including PayPal web and knowledge-base content, Autodesk, SolidWorks and Adobe software, and technical material for HP and Intel.
This was where the basic habits formed: terminology had to remain stable across large products, references had to be checked rather than guessed at, and clear Polish mattered just as much as technical correctness.
It was also my introduction to localization as a production process rather than a collection of individual translations: multiple linguists, shared terminology, recurring content, review stages, legacy decisions, and source material that did not always explain itself.
2009–2013 · From translation to voice
Press-release work for Western Digital and NVIDIA brought more direct client communication, copywriting, and transcreation into the mix.
The problem was no longer always what does this sentence mean? Increasingly, it was what is this sentence supposed to achieve?
In 2012, work with Localsoft began a long involvement with game localization. Projects covered RPGs, action games, resource-management games, storefront copy, community material, and subtitles.
Work around World of Tanks, World of Warplanes, and World of Warships combined in-game text with historical articles, video subtitles, review, reference compliance, and translator test assessment.
Games made context harder to ignore. A line might belong to a character, an interface element, a tutorial, a historical description, or a piece of dialogue whose meaning depended on something the string itself never mentioned.
That gradually shifted my work from translating individual strings toward maintaining voices, terminology, and continuity across an entire product.
2014–2018 · Language ownership and creative production
Work with Yamagata Translations broadened the technical range further: software, manuals, documentation, and instructional and promotional subtitles across automotive, logistics, video, printing, and design.
Craft Translation, beginning in 2015, moved much further into creative production: advertising, taglines, brand guides, social campaigns, scripts, apps, websites, and recording work.
That meant adapting copy around timing, rhythm, visuals, cultural assumptions, performance, and client intent rather than treating the source sentence as something that had to survive intact.
The commercial and transcreation side of that work eventually covered a much broader range of brands and industries; some of them are listed on my transcreation page.
From 2016, I also served as Polish Language Owner for Oracle legal content. The role added terminology and style maintenance, client coordination, quality review, recruitment support, and responsibility for keeping work consistent with formal requirements across a continuing body of material.
2019–2023 · UX, production scale, and missing context
Wix projects in 2020 brought UX writing and localization to the foreground: knowledge-base articles, tooltips, feature descriptions, and system messages that had to explain sometimes complicated behaviour in very little space.
Other work during the period included enterprise security software and an online ordering and logistics platform.
At Craft, assignments increasingly involved production supervision, linguistic leadership, cultural QA, and adapting copy during live recording sessions. The work was no longer confined to the translated text itself; it extended into how that text behaved when spoken, displayed, reviewed, and used.
Beginning in 2020, I also worked with Localsoft on multiple NDA-protected AA and AAA games for PC, VR, and consoles.
These were production-scale projects involving dialogue and UI datasets, narrative and character references, terminology and legacy-text continuity, structured tags and variables, CAT-tool workflows, review, and screenshot-supported linguistic QA. Some contained tens of thousands of strings and extensive context packs; titles and source material remain confidential.
The recurring difficulty was increasingly obvious: the translation environment often knew much less about a string than the game did.
Who is speaking? Who are they speaking to? What happened immediately before this line? Is a variable a person, an item, or a place? Does the character have a defined gender? Is this text shown in a menu, spoken aloud, or assembled dynamically at runtime?
Part of the job therefore became recovering context that the localization pipeline had discarded.
That led to more team leadership, mentoring, style maintenance, and small tools for problems such as gender variants, character limits, terminology, and incomplete context.
By 2023, those systems problems were becoming a central part of my work rather than occasional obstacles around the edges of it.
2024–present · Localization systems
I was contracted as a Polish Localization Writer for the unreleased Kenshi 2, which brought many of those threads together. Its reactive dialogue places unusual demands on Polish: grammatical gender and number interact with speaker and listener properties, faction identity, world state, reusable text fragments, and other variables. Text that looks like a single source line can represent a large family of possible sentences.
The work therefore involves not only writing Polish dialogue designing and maintaining the linguistic rules that allow those sentences to remain grammatical, coherent, and in character.
That exposed a practical problem: conventional localization software could store the strings, but it could not adequately show what those strings might become.
I therefore built a game-specific authoring and QA environment around the dialogue system itself. It retrieves context directly from the source data, resolves grammatical and conditional variants, previews generated sentences, and checks combinations that would otherwise be impractical to inspect by hand.
Instead of translating and reviewing isolated templates, translators can work much closer to the sentences players may actually see — while still tracing those sentences back to the rules and source strings that produced them.